Abstract
Post-glacial climate, vegetation and fire history were reconstructed from a sediment record from Begbie Lake, British Columbia, Canada, located in a municipal water supply area servicing > 350,000 people. Watershed managers have identified wildfire as a threat to water supply and seek to understand how vegetation and fire have varied through time with climate. In the cold late-glacial, open Pinus woodlands, periodically disturbed by fire, transitioned to mixed conifer forests subject to high-severity fire. The early Holocene is of interest to watershed managers because climate was warmer and drier than present. During this interval, low streamflow, abundant fire-adapted taxa, elevated background charcoal and regional increases in biomass burning indicate that fire seasons were longer and that fire was an important disturbance mechanism. Climate moistened in the mid Holocene, facilitating canopy closure and decreased fire disturbance. However, surface fires prevailed in Quercus ecosystems, which were expanding locally. Charcoal increased between 6180–2500 cal yr BP as climate further cooled and moistened, likely reflecting human activity and/or increased climate variability. Modern conditions arose within the last few millennia, impacted most recently by European settlement. In combination with paleoclimate modelling, modern management practices and forecast simulations, the Begbie Lake record informs about ecosystem changes within the watershed, yielding insights for management.
Introduction
Climate change is driving a general increase in wildfire frequency and area burned (Abatzoglou and Williams, 2016; Flannigan et al., 2005, 2009; Gillett et al., 2004). In Canada, wildfire occurrence, size and severity are projected to increase by the end of the century (Flannigan et al., 2005), with feedbacks further exacerbating or diminishing responses (Brown and Giesecke, 2014; Girardin et al., 2013; Lynch et al., 2002; Price et al., 2013; Wang et al., 2016). Although fire is a natural and essential process in wildlands, where it cycles nutrients, creates vegetation mosaics, increases biodiversity and potentially reduces the impact of insects and disease, it also poses a serious threat to communities and industry, necessitating investment aimed at decreasing the risk and impact. Fires in the wildland–urban interface, which is defined as the boundary separating developed areas from undeveloped regions (Johnston and Flannigan, 2018), threaten public safety, property, infrastructure and other critical investments, including those related to natural resources. Thus, several national-scale initiatives have been developed to mitigate current and future wildfire threats, including the Canadian Wildland Fire Strategy (McFarlane, 2006; Taylor et al., 2006; Wildland Fire Management Working Group, 2016). Other programmes, such as FireSmart, aim to enhance interagency cooperation to increase public awareness and decrease risk associated with fire disturbance (Partners in Protection, 2003).
Select examples of recent, well-known wildland–urban interface fires in Canada include the 2003 Okanagan Mountain Park Fire, Kelowna, British Columbia (BC; Filmon et al., 2004); the 2011 Flat Top Complex Fire, Slave Lake, Alberta (KPMG, 2012); and the 2016 Horse River Fire, Fort McMurray, Alberta (KPMG, 2017). In the first example, hot and dry conditions enabled a 25,600 ha fire to burn into the southern portion of the City of Kelowna, causing the evacuation of 27,000 people and destruction of 239 homes. Similarly, the Slave Lake fire destroyed 730 homes and numerous businesses and caused the evacuation of ca. 15,000 people (KPMG, 2012). In the case of the Horse River fire, high spring temperatures, low relative humidity and low winter snowpack facilitated a > 500,000 ha fire that destroyed 1958 structures and forced the evacuation of 88,000 residents. The total cost of this fire is estimated at > $8.9 billion, making it the costliest natural disaster in Canadian history (KPMG, 2017). Post-fire dynamics were also important, with concerns raised about rain-induced ash transport contaminating the Fort McMurray water supply.
The impacts of wildfire on surface water quality are well-studied (Neary et al., 2005; Tiedemann et al., 1979). A comprehensive review by Smith et al. (2011) concluded that enhanced post-fire erosion can contaminate water supplies by increasing levels of suspended sediments, nutrients (e.g. N, P) and potentially harmful elements (e.g. As, Pb, Hg). Other elements, including Cl, Fe, Mn, Na and Zn can impact the aesthetics of the water, variously affecting taste and colour. Likewise, a study of seven burned watersheds by Emelko et al. (2011) found that contaminants, including suspended sediment, dissolved organic carbon, total P and Hg were elevated in burn site discharge for 2–4 years after disturbance. The authors warn that conventional catchment management strategies leave water purveyors ill-prepared to treat contaminated run-off and emphasize the need for improved source water protection. Despite this growing awareness, a review by Bladon et al. (2014) asserts that effective watershed management requires a better understanding of wildfire impacts on water quality, especially with regard to drinking water supply. Consequently, watershed managers are often faced with the question of whether wildfires will create conditions that exceed established water quality standards (Neary et al., 2005).
The Sooke Lake Reservoir (SLR) is a dammed natural lake that supplies municipal water to > 350,000 people on southern Vancouver Island, BC, Canada (Capital Regional District (CRD), 2012). Approximately 98% of the forested catchment area of SLR is owned and managed by the CRD, a level of local government that provides regional services. While the forested lands within the Sooke Lake Watershed (SLW) yield high-quality drinking water at comparatively low cost, they are susceptible to changes in climate and fire disturbance (CRD, 2012, 2015; Smith et al., 2011). For example, numerous paleofire studies have shown that disturbance regimes in BC have changed through time in response to changes in climate, vegetation and human action (Allen, 1995; Brown and Hebda, 2002a, 2002b; Brown et al., 2006, 2017; Gavin et al., 2003; Hallett et al., 2003; Hebda, 1995; Heinrichs et al., 2004; Lucas and Lacourse, 2013; Sanborn et al., 2006). Recently, according to the BC Wildfire Service (https://www2.gov.bc.ca/gov/content/safety/wildfire-status), annual area burned in BC from 2007–2016 ranged from 13,000-370,000 ha. In contrast, area burned during the 2017 and 2018 fire seasons was ca. 1,200,000 and 1,300,000 ha, respectively, with suppression costs in the former totalling ca. $568 million. Considering the future, a simulation of provincial ecosystem climate niches indicates that the vegetation climate space will change in BC over the coming decades (Wang et al., 2012). Likewise, fire simulations indicate that severity ratings and annual area burned will also increase (Haughian et al., 2012; van der Kamp et al., 2013; Wang et al., 2016). In response, the CRD has identified wildfire as a major threat to water supply (CRD, 2012, 2015).
Given the anticipated changes in climate, vegetation and fire disturbance, the CRD seeks to better understand wildfire disturbance dynamics and impacts within SLW. Consequently, this study uses paleoenvironmental records to characterize the long-term relationships among climate, vegetation and fire disturbance in the watershed. The study also seeks to quantitatively reconstruct precipitation changes through time using a pollen-precipitation transfer function and integrate those data into a water balance model to reconstruct streamflow for two catchments, Rithet and Judge, in the watershed. Of particular significance is the early Holocene interval (11,700–7000 calendar years before present; cal yr BP). While this interval does not provide an exact analogy for the future due to differences in forcing mechanisms (insolation vs. greenhouse gas emissions) and ecological conditions (natural systems vs. introduction of invasive species, fragmented landscapes, environmental pollution and degradation), it is of interest because the climate was warmer and drier compared with present, with associated changes in vegetation and fire disturbance (Hebda, 1995, 1998). In addition to revealing temporal climate–vegetation–fire interactions, the long-term fire record will also be examined for evidence of past anthropogenic burning within the watershed. The results will ultimately be integrated into the CRD’s climate adaptation strategy, highlighting the applicability of paleoecological research to resource management (Brown et al., 2018). The approach developed here could be generalized to other water supply areas across Canada and internationally.
Study site
SLR is located on southern Vancouver Island, BC, Canada (48.55° N, 123.71° W, 187 m above sea level; Figure 1). With a surface area of 7.35 km2 and a holding capacity of 92.7 million m3, this elongate lake is the primary water supply for the residents of greater Victoria, the capital city of BC, and surrounding municipalities. The reservoir is drawn on year-round and has a mean water residence time of about 2 years (Barraclough, 1995). Its naturally oligotrophic, low-turbidity waters receive no filtration before delivery to customers, which makes this water source particularly susceptible to fire-induced contamination (CRD, 2012). SLR is situated within ca. 70 km2 of forested watershed defined by two long ridges running north–south on either side of the reservoir (CRD, 2015; Figure 1), yielding a surface area to catchment ratio of 1:10 (Werner et al., 2015). The area experiences warm, dry summers and mild winters with intermittent snowfall. Precipitation is relatively light, averaging 1640 mm annually (Werner et al., 2015).
Series of progressively zooming nested location maps, as per the red rectangle in each panel. (a) Regional location map showing location of British Columbia (black) within Canada. (b) Map showing location of the Greater Victoria Water Supply Area (GVWSA; black filled polygon) on Vancouver Island. Also shown are regional biogeoclimatic zones, including the Coastal Douglas Fir (CDF), Coastal Western Hemlock (CWH), Mountain Hemlock (MH) and alpine tundra (AT) zones. (c) The three watersheds of the GVWSA, including the Leach River (LRW), Sooke Lake (SLW) and Goldstream (GSW) watersheds. The Sooke Lake Reservoir (SLR; blue shaded) is located within SLW, with 0 m bathymetry representing the shoreline when the reservoir is at full capacity. CWH xeric maritime (CWHxm) and CWH moist maritime (CWHmm) biogeoclimatic subzones are also shown, legend in (b). (d) Northern portion of SLR, showing the Rithet (R) and Judge (J) creek catchments, drainages and gauging stations (grey dots with black outline) relative to Begbie Lake (shown in red box). (e) Begbie Lake coring location (black dot).
Biogeoclimatic zones are used in BC to delineate broad ecosystems based on regional climate and climax vegetation (Meidinger and Pojar, 1991). These communities are further divided into subzones according to levels of precipitation (xeric, moist and very wet) and degree of continentality (maritime to hypermaritime). The Coastal Western Hemlock (CWH) biogeoclimatic zone prevails within SLW, which is primarily comprised of the xeric maritime (CWHxm) subzone. A small area of moist maritime (CWHmm) subzone exists on the western margin of the watershed. Although fire is not the dominant natural disturbance mechanism in moister CWH subzones outside the study area to the west, its importance increases in drier settings like SLW. It should be further noted that during the 2018 fire season, several lightning-ignited fires burned large patches of CWH forest on northern Vancouver Island under dry summer conditions, revealing the vulnerability of CWH forest to fire disturbance.
Timber harvesting and other activities within SLW over the last 100 years have contributed to the establishment of younger age-classes (CRD, 2015). Following the reduction in historical logging and the introduction of fire suppression more than a century ago (MacDonald, 1929), forest fuel loads have increased. The canopy is now thicker and more enclosed relative to the mid-20th century, with a less diverse mosaic of stand ages and fuels (CRD, 2015). Flammable invasive species, such as Cytisus scoparius, are present in disturbed areas and along some roads.
Due to the challenges related to coring directly within SLR in > 75 m water depth without disturbing water supply, an alternate basin was selected for sampling. Begbie Lake (Figure 1; 48.58° N, 123.68° W) is a small water body (0.3 ha) situated approximately 400 m north of SLR. It was an isolated lake prior to 2002, but a recent raising of SLR connected the lake to the reservoir via a channel. Begbie Lake is relatively shallow, enabling manual retrieval of a long sediment core. Critically, its proximity to SLR ensures that the two lakes share common climate, vegetation and fire histories. The vegetation immediately around Begbie Lake consists of multiple age classes. The tree stratum is dominated by Pseudotsuga menziesii, with lesser amounts of Tsuga heterophylla and Thuja plicata. Scattered Pinus contorta, Acer macrophyllum, Alnus rubra and Abies grandis occur around the basin. The understory contains small T. heterophylla trees as well as Gaultheria shallon, Mahonia nervosa and Polystichum munitum coupled with mosses. Aquatic vegetation includes Myrica gale, Cyperaceae, Potamogeton and Nuphar.
Several drainages empty into SLR, including Rithet Creek to the northwest and Judge Creek to the north (Figure 1). About 25% of the water entering SLR is supplied by Rithet Creek, a north–south oriented fourth order catchment encompassing 18.54 km2 west of Begbie Lake. It is characterized by a pluvial hydroclimatic regime with average precipitation of 1917 mm yr−1. Streamflow in Rithet Creek shows marked intraannual variability, with highest monthly discharge occurring in January (average: 270 mm month−1) during the November–February wet season, and lowest discharge in August (average: 4 mm month−1) during the June–September dry season. During droughty summers, flow rates can even be lower. Most of the Rithet catchment is within the CWHxm subzone, though the highest elevations of the southwest portion is within the CWHmm subzone. The Rithet Creek gauging station is located at the southern end of the catchment (48.57 °N, 123.71 °W) and has been gathering discharge data continuously since AD 1994. In contrast, the Judge Creek drainage is a third-order catchment that is both smaller (7.93 km2) and slightly drier (average annual precipitation: 1561 mm) relative to the Rithet Creek catchment. This pluvial basin is located northeast of Begbie Lake, exclusively within the CWHxm subzone. The Judge Creek gauging station is located at the southern end of the catchment (48.59 °N, 123.67 °W), likewise gathering discharge data continuously since AD 1994. The annual discharge cycle is comparable with that at Rithet Creek, with highest monthly discharge in January (average: 232 mm month−1) and lowest monthly discharge in August (average: 0.24 mm month−1).
Methods
Using an inflatable Zodiac raft-based coring platform, a sediment record was collected from Begbie Lake (Figure 1). The sediment core was recovered from near the centre of the lake basin in 827 cm water depth. Two coring devices were used to retrieve the record. A 50-cm long Kajak corer (Kajak et al., 1965) was used to collect the uppermost unconsolidated sediment, which was subsequently sectioned into 0.5 cm intervals in the field. A modified Livingstone piston corer (Wright, 1967) was used to recover sediment below 30 cm depth. Because this process is incremental, two overlapping cores were collected (horizontal offset = 1 m, vertical offset = 50 cm), ensuring recovery of a continuous record.
The sediment cores were photographed in the laboratory and a chronology established using both lead (Pb-210) and radiocarbon (14C) dating (Table 1). The topmost sediment samples (0–52.5 cm depth) were analysed by Flett Research Ltd., Winnipeg, for Pb-210 dating. According to their report, the core showed an irregular, but approximately exponential decrease in Pb-210 activity as a function of depth, with similar Pb-210 and Ra-226 activity at depth, suggesting background levels of Pb-210 may have been achieved. Noting, however, that the exponential decay of Pb-210 suddenly terminates at 27.25 cm depth, the report suggests that the Pb-210 inventory is actually incomplete. Consequently, Flett Research Ltd. determined the age and accumulation rate of sediment using a constant rate of supply (CRS) model that was calibrated against a linear regression model of unsupported Pb-210 activity versus cumulative dry weight (g cm−2). This allowed the CRS model to be used over 0–22.25 cm depth, with predicted ages greater than 80 years being approximations only. Thus, only ages less than 80 years were used in the development of the age-depth model. In addition, plant macrofossils and bulk samples were sent to the A.E. Lalonde AMS Laboratory, University of Ottawa, for 14C dating. Radiocarbon dates were calibrated using Calib 7.1 (Stuiver et al., 2019). A 1-cm thick layer of Mazama ash was observed in the sediment core at 406 cm depth, providing another chronological control point (Egan et al., 2015). Given that the ash reflects a discrete eruption event, its thickness was subsequently removed from age-depth modelling. The age-depth model was established using Clam 2.1 (Blaauw, 2010), whereby a smoothing spline with 95% confidence intervals was fit to the chronological control points.
Chronological control points for Begbie Lake.
N/A: not available.
from Egan et al. (2015).
To reconstruct vegetation and climate history, 1 cm3 sediment subsamples were extracted every 10 cm, spiked with an exotic Lycopodium tablet (20,848 ± 1,546 spores tablet−1; Department of Quaternary Geology, University of Lund, batch no. 1031) and processed for pollen and spores (herein ‘pollen’) using standard techniques, excluding hydrofluoric acid treatment (Moore et al., 1991). Pollen slides were made by homogenizing 35 µl of residue, measured using a single-channel pipette, with 15 µl of melted glycerin jelly, followed by the application of a glass cover slip. Slides were counted using a Leica DM4000 B LED compound microscope at 400–630× magnification. A reference collection from the Royal British Columbia Museum and published keys (Moore et al., 1991) aided identification. In addition to pollen percentage, pollen accumulation rate (grains cm−2 yr−1) was calculated by multiplying pollen concentration (grains cm−3), determined using the number of Lycopodium spores encountered, by the vertical accretion rate (cm yr−1). A previously developed Pseudotsuga (Douglas-fir) to T. heterophylla (western hemlock) pollen index (DWHI; Brown and Schoups, 2015; Brown et al., 2006) was used to quantitatively reconstruct changes in Holocene precipitation. The upper DWHI value from the Begbie Lake core provided an additional data point that was added to the existing regression model (Brown and Schoups, 2015), slightly changing the regression parameters. The DWHI ratio is based on the fact that Pseudotsuga favours xeric settings whereas T. heterophylla prefers hydric environments. The DWHI regression model was developed using modern pollen samples from lake surface sediment and gridded precipitation from PRISM (Daly et al., 1994, 2008; Hamann and Wang, 2005). Temporal estimates of Holocene precipitation were estimated using downcore DWHI values. The pollen diagram was generated using Tilia 1.7.16 (Grimm, 1990), whereby stratigraphically constrained cluster analysis was used to establish pollen zones (Grimm, 1987).
Charcoal is produced by pyrolysis and is a product of incomplete biomass combustion, thus, providing direct evidence of burning. It is dispersed by atmospheric transport and surface flow during and after fire events, with some entering lake basins where it is preserved (Brown and Power, 2013; Conedera et al., 2009). To reconstruct fire history, 3 cm3 subsamples were collected from the sediment core in contiguous 1 cm intervals using a modified syringe. The subsamples were passed through a 150 µm sieve to remove the fine-grained sedimentary matrix and other small particulates. For each subsample, the coarser fraction was retained and subsequently placed in a gridded petri dish in distilled water where it was examined using a Leica MZ-7 stereomicroscope at 40× magnification. Fragments were classified as charcoal, if they were black, opaque and brittle, with visible cellular structure. The charcoal time series was decomposed using CharAnalysis 0.9 (Higuera et al., 2010). Charcoal concentration (fragments cm−3) was interpolated to the median sample resolution of the core (21 years) and converted into charcoal accumulation rate (CHAR, fragments cm−2 yr−1) by multiplying by the vertical accretion rate. The CHAR series was statistically decomposed into background and peak components using a Lowess smoother with a 750-year window. The background component reflects long-term changes in charcoal production, transport and sediment mixing. The peak component is comprised of fire-induced signal and non-fire related noise. For every sample in the record, a Gaussian mixture model with a 95th percentile separation threshold was used to separate signal from noise, thus generating a record of fire events through time. A signal-to-noise index (SNI) was used to evaluate if the record was suitable for peak detection (Kelly et al., 2011). The trend in Holocene fire frequency was generated using a smoothing window, whereby the total number of fires in each 1000-year period were summed and smoothed using a Lowess smoother. A similar process was used to determine trends in fire return intervals (FRI; Higuera, 2009).
During a fire, non-ferrimagnetic iron minerals in soils are converted to strongly ferrimagnetic oxides at a threshold of 100–200 °C (Conedera et al., 2009; Gedye et al., 2000). This magnetized soil can be eroded from the burned area and deposited in lake sediments, forming a magnetically distinct layer (Conedera et al., 2009) that can be identified by measuring sediment magnetic susceptibility. The core was analysed for magnetic susceptibility in contiguous 1 cm increments using a Bartington MS3 metre and MS2E core logging sensor, with measurements logged using Bartsoft for PC v4. Temperature drift corrections were automatically applied every five samples. Holocene streamflow was reconstructed for Rithet and Judge creeks using a calibrated water balance model for each catchment, with model development, validation and assumptions described in detail in Brown and Schoups (2015). Briefly, the model accounts for temporal changes in annual precipitation by applying the DWHI-precipitation transfer function to down-core changes in DWHI values from the Begbie Lake core. Temporal changes in average monthly, basin-scale precipitation were computed by distributing annual precipitation over the months, using modern and past precipitation fractions (6000 and 11,000 cal yr BP; Bartlein et al., 1998; Kutzbach et al., 1998). Similarly, temporal changes in evapotranspiration were estimated using Holocene changes in solar radiation (Berger and Loutre, 1991) and simulated temperature, with the latter adjusted similarly to precipitation. The model then accounts for partitioning effective rainfall into streamflow, evapotranspiration and soil water storage components, as a function of water storage in the catchment, for each month in 500-year intervals.
Results
The Begbie Lake sediment core was 633 cm long, with the oldest date of 13,800 cal yr BP at 621.5 cm depth (Figure 2). The core was predominately comprised of gyttja, with Mazama ash at 405–406 cm depth. Towards the bottom of the core, gyttja gradually yields to grey clay (614–624 cm), with clay and sand comprising the basal sediment from 624–633 cm depth. Six pollen zones were identified in the Begbie Lake core (BL-1 to BL-6; Figures 3 and 4), with the top zone partitioned into two subzones (BL-6a and BL-6b). The oldest zone, BL-1 (614–633 cm; ⩾ 13,510 cal yr BP), was dominated by Pinus subgenus Pinus (diploxylon) pollen. Zone BL-2 (564–614 cm; 11,560-13,510 cal yr BP) contains a mixed conifer pollen assemblage consisting of Pinus, Picea, Abies and Tsuga mertensiana together with Populus, Alnus and Salix. Early Holocene zone BL-3 (424-564 cm; 8020–11,560 cal yr BP) was characterized by increasing amounts of Pseudotsuga, Rosaceae, Poaceae and Pteridium. Rare occurrences of Selaginella are also noted. Alnus and Salix remain as important components of the vegetation. Zone BL-4 (324–424 cm; 6180–8020 cal yr BP) contains increasing amounts of T. heterophylla, Cupressaceae (likely T. plicata) and Cyperaceae. Pseudotsuga, Alnus and Pteridium decreased relative to the preceding zone, whereas Rosaceae remains similar to BL-3. Quercus reaches its Holocene maximum values in the upper portion of BL-4 above Mazama ash. Cupressaceae and Cyperaceae increased in Zone BL-5 (135–324 cm; 2500–6180 cal yr BP). T. heterophylla and Pseudotsuga with Alnus and Rosaceae were also notable. Quercus occurs throughout the zone. T. mertensiana and Lysichiton recur intermittently. BL-6a (25–135 cm, 260–2500 cal yr BP) was characterized by increasing Abies, Pseudotsuga, T. heterophylla, Myrica and Lysichiton. Cupressaceae and Alnus decreased throughout the zone. At the top of the core in Zone BL-6b (0–25 cm, 64–260 cal yr BP), Alnus as well as several non-arboreal types increased in abundance.
(a) Pb-210 age–depth relationship determined using a constant rate of supply model (black dots). Dates greater than 80 years (x symbols) are approximate and were not used in age–depth modelling. (b) Post-glacial age–depth model (black line) with Pb-210 at top (black box) and calibrated ranges (plotting points). Also shown are 95% confidence intervals (grey shaded area).
Begbie Lake selected pollen and spore percentage diagram, including zones (BL-1 to BL-6b) as defined by CONISS.
Begbie Lake selected pollen and spore accumulation rate diagram.
Regarding fire disturbance, the SNI of Kelly et al. (2011) was above 3.0 for most of the core, with a global index of 4.6, indicating good separation of signal from noise (Figure 3). There are a few instances where the index briefly decreased below 3.0, attaining low values of 2.6 around 1600 cal yr BP, 2.9 at ca. 9200 and 2.8 at ca. 11,100 cal yr BP. A total of 59 fire events were identified in the core and oscillations in background charcoal were observed. Four fire events were identified in basal Zone BL-1, with FRIs of 170 years. Background charcoal increased in Zone BL-2 (average: 1.4 fragments cm−2 yr−1), as do the number of fire events (6). FRIs were more irregular compared with the preceding zone, ranging between 80 and 780 years. In BL-3, background charcoal was relatively high, averaging 1.8 fragments cm−2 yr−1. Further, 14 fire events were identified in the zone, with longer FRIs (340–610 year range; average: 470 years) between 11,560 and 9500 cal yr BP compared with 60–270 years between 9500 and 8020 cal yr BP (average: 150 years). Many of these events were characterized by small peak magnitudes. In BL-4, both background charcoal (average: 1.2 fragments cm−2 yr−1) and the number of fire events (8) decreased relative to BL-3. Longer FRIs (270–360 years) occurred between 8020 and 6500 cal yr BP compared with 40–130 years between 6500 and 6180 cal yr BP. Background charcoal increased notably in BL-5, averaging 2.2 fragments cm−2 yr−1. The number of fire events likewise increased to 16, with the FRI ranging between 60 and 440 years and averaging 225 years. Background charcoal decreased in BL-6a to 1.2 fragments cm−2 yr−1 and fewer (10) fire events were identified, with FRIs ranging between 60 and 420 years. Background charcoal increased in the topmost zone (BL-6b), averaging 1.3 fragments cm−2 yr−1, with highest values occurring at the top of the core. One fire event was identified in BL-6b.
Magnetic susceptibility values generally decreased through time, with the highest values occurring near the bottom of the core (Figure 3). In BL-1, the highest and most variable magnetic susceptibility values (4.55–22.86 × 10−4 SI) were observed before 13,250 cal yr BP at 626–633 cm depth. Thereafter, values were generally lower (average: 1.01 × 10−4 SI) and with moderate variability (0.44–1.64 × 10−4 SI) compared with the bottom-most samples. Intermediate magnetic susceptibility values continue to prevail throughout BL-2 (average: 1.06 × 10−4 SI) and the lower portion of BL-3 (average: 0.90 × 10−4 SI). Over this interval, the two largest peaks (8.09 and 4.82 × 10−4 SI) occurred at 12,050 and 10,430 cal yr BP (577 and 529 cm depth, respectively). Magnetic susceptibility values and variability decreased further in the upper portion of BL-3 from 521–424 cm depth, averaging 0.22 × 10−4 SI. Consistently low magnetic susceptibility values (average: –0.01 × 10−4 SI) occur in BL-4, with the exception of a Mazama ash-induced peak at 406 cm. A notable increase in magnetic susceptibility was observed in BL-5 from 5850–3900 cal yr BP (206–309 cm depth; average: 0.26 × 10−4 SI), after which lower values recur (average: –0.03 × 10−4 SI). This general trend continued until the top of the core, with BL-6a and -6b values averaging 0.02 × 10−4 SI. Some magnetic peaks appear to temporally coincide with charcoal peaks, potentially reflecting higher severity burns (Dunnette et al., 2014; Figure 3).
Adding an additional data point (0.55 DWHI and 1386 mm) from the topmost sample of the Begbie Lake core to the existing DWHI-annual precipitation regression (Brown and Schoups, 2015) did not profoundly change the regression. Following data transformation, the best fit parameters of the new linear regression are a = 737 and b = 1.63, with a RMSE (root mean square error) = 534 mm, which are similar to the previous regression (744, 1.64 and 535 mm, respectively). The regression is annual precipitation = 737.1563 × exp(1.632 × DWHI). The best estimates of precipitation as a function of DWHI occur between 740 and 3770 mm, which was similar to the previous regression (750–3800 mm). The precipitation reconstruction revealed that precipitation was relatively low (ca. 830 mm) at 10,000 cal yr BP. Thereafter, values trend upwards until 4000 cal yr BP (ca. 1600 mm). Precipitation decreases somewhat until 2000 cal yr BP (ca. 1300 mm), increasing thereafter until present day (ca. 2100 mm).
The water balance models for the Rithet and Judge catchments were forced with reconstructed Holocene precipitation using the DWHI regression, resulting in reconstructed monthly discharge (Figure 5). The model parameters for both catchments were the same as those obtained by Brown and Schoups (2015) for the nearby Koksilah River basin, located north of SLW. The water balance models correctly reproduce historical discharge data for Rithet and Judge creeks, with RMSE values of 13.9 and 13.3 mm month−1, respectively (Figure 5a and 5b). Similar trends in Holocene discharge were obtained in both catchments (Figure 5c and 5d). At 10,000 cal yr BP, discharge was significantly lower than it is now, with lengthened dry-season conditions prevailing in both catchments. Over the next 6000 years both catchments gradually became wetter resulting in increasing discharge, especially during the wet season. The wetting trend reversed between 4000 and 2000 cal yr BP, resulting in discharge values comparable with those at 8000 cal yr BP, with wet-season discharge less than half of what is currently observed. During the last 2000 years, renewed moistening resulted in modern discharge rates.
Simulated versus observed historic mean monthly streamflow for (a) Rithet Creek and (b) Judge Creek. Simulated Holocene streamflow for (c) Rithet Creek and (d) Judge Creek. Months 1 to 12 represent January to December. Note that the 4000 cal yr BP line is overlain by the nearly identical 6000 cal yr BP line.
Discussion
Deglaciation, primary succession and late-glacial fire regime
While portions of Vancouver Island were likely ice-free during the Cordilleran Glaciation (Brown and Hebda, 2003; Hebda, 1997a), many regions were covered in ice or by glaciers. In response to changing orbital geometries, the Cordilleran Ice Sheet began to ablate ca. 14,000 cal yr BP (Booth et al., 2003), leaving behind a landscape covered in drift and meltwater deposits together with kettle and glacially scoured lakes. Silt and clay deposits characterized by markedly high magnetic susceptibility values at the bottom of the Begbie Lake core (BL-1) reflect this open, erodible landscape. The preponderance of diploxylon Pinus pollen (Figures 3 and 4), suggests that pioneering pine trees, likely P. contorta (Brown and Hebda, 2003; Brown et al., 2008), quickly colonized the deglaciated landscape. Given that Pinus is overrepresented in the modern pollen rain (Allen et al., 1999; Hebda and Allen, 1993), it is hypothesized that open-canopy pine-dominated woodlands prevailed at this time. Lesser amounts of Alnus pollen, suggest that nitrogen-fixing alder occupied openings, possibly facilitating a period of primary succession and pedogenesis. Alnus and Salix also likely grew at the lake margins and in other riparian settings. Regionally, the vegetation supported megafauna, including Bison antiquus, with butcher marks on fossil bones illustrating a human presence on the landscape (Kenady et al., 2011; Leopold et al., 2016; Wilson et al., 2009).
Although charcoal is rare in the bottom-most part of BL-1 (>13510 cal yr BP), it quickly increases in abundance further up the zone at 626 cm depth. Charcoal was similarly noted in a late-glacial (Pinus zone) deltaic sequence located about 100 km northwest of Begbie Lake (Terasmae and Fyles, 1959). Together, these records imply that fires may have been burning under cold, dry climatic conditions, with a mean FRI of 170 years. Thus, even though the Pinus woodlands were open, climate-mediated slow rates of biomass decay coupled with long intervals of time between fires resulted in sufficient fuel accumulation for periodic combustion. These results, however, differ from other regional investigations that used a coarser sampling resolution, finding little or no charcoal in the Pinus zone (Brown & Hebda, 2002b, 2003). In those studies, it was suggested that the apparent lack of fire was related to the unavailability of an ignition source. Given that the land surface was highly erodible at this time, another possibility is that the aforementioned differences in charcoal deposition relate to markedly different rates of sediment input to the basins (i.e. erosion), diluting or enriching charcoal concentrations. This issue is compounded by the lack of available basal radiocarbon dates, making interpretation difficult. It would be beneficial if these different interpretations could be reconciled because they influence the ecological interpretation of the Pinus zone and its disturbance regime. For example, the coastal variety of Pinus (P. contorta var. contorta) is not as dependent on fire as the interior variety (P. contorta var. latifolia; Brown and Hebda, 2002b).
The Pinus woodlands were replaced by mixed conifer forest (BL-2; 11560-13510 cal yr BP) as climate cooled and moistened (Figures 3 and 4). The pollen record indicates that while Pinus decreased in prevalence, Picea, Abies, T. mertensiana and, to a lesser extent, Populus, increased in abundance. The discovery of Abies lasiocarpa and Picea engelmannii needles in nearby contemporaneous sediment (Allen, 1995; Barnosky, 1981; Heinrichs et al., 2002) suggests that the Abies pollen was produced by sub-alpine fir, while the Picea pollen was derived from Engelmann spruce. Both species are climax co-dominants that are tolerant of shade (Coupé et al., 1991; Heinrichs et al., 2002), suggesting that the forest canopy was closed. Today they occupy areas characterized by cold winters, cool summers and thick snow packs. Moreover, they are thin-barked with flammable foliage and highly sensitive to fire disturbance, where even low-severity fires can induce mortality. Their presence implies that fire disturbance was infrequent, likely producing even aged-stands. Indeed, the charcoal record reveals that the FRI was generally long, though variable, ranging from 70 to 730 years (averaging 340 years; Figure 3). Moderately high and variable magnetic susceptibility values throughout the zone may reflect combined effects of continued deposition of unconsolidated surface sediment coupled with episodic high-severity fires in the watershed, elevating the magnetic signal. In addition to responding to disturbance-induced canopy openings, Alnus and Salix likely continued to grow around the lake basin.
Fire disturbance and adapted species in the warm, dry early Holocene
Higher-than-present summer and lower-than-present winter insolation characterized the Northern Hemisphere during the early Holocene interval (Cooperative Holocene Mapping Project (COHMAP), 1988). These differences, together with the development of a subtropic high pressure system in the northeastern Pacific Ocean, resulted in greater seasonality, warmer summer temperatures and overall drier conditions with enhanced summer drought (Brown et al., 2006; COHMAP, 1988; Hebda, 1995). While summers in the interior of BC were 2–4 ºC warmer-than-present (Rosenberg et al., 2004), the coastal regions were probably only 1–2 ºC warmer (Hebda, 1995; Mathewes and Heusser, 1981). DWHI values from the Begbie Lake core indicate that precipitation was generally low in the early Holocene (830–1420 mm from 7000 to 10,000 cal yr BP), consistent with other regional sites (Brown et al., 2006). As with other watersheds (Brown and Schoups, 2015), streamflow in the Judge and Rithet catchments were characterized by longer low-flow seasons (e.g. April–October; Figure 5) and overall lower discharge in both summer and winter relative to present. As such, it is posited that the fire season was likewise of longer duration. Under these conditions, fire-adapted P. menziesii arrived and expanded on Vancouver Island, becoming particularly abundant within SLW (BL-3; 8020-11560 cal yr BP). Pteridium, Alnus, Rosaceae, Poaceae and, to a lesser extent, Selaginella likewise increased in abundance, whereas conifers present in the late-glacial became less common.
P. menziesii is generally shade intolerant and is often a pioneering species after disturbance such as fire (Farrar, 1995; Nuszdorfer et al., 1991). Without disturbance, it will eventually yield to shade-tolerant T. heterophylla and T. plicata in mesic sites and A. grandis in drier settings. The Alnus pollen likely represents A. rubra, an early seral and riparian species. Abundant Pteridium spores indicate that bracken fern was important in the understory. Pteridium is a fire-promoting species that can survive in a high-frequency disturbance regime (Crane, 1990) since it has deeply buried rhizomes that are well suited for colonizing open areas sterilized by fire. The increase in rosaceous pollen suggests somewhat open forest conditions that enabled the rose signal to emerge. Poaceae pollen averages 2% of the total pollen in BL-3. According to Allen et al. (1999), open grasslands have between 5-20% grass pollen, implying that expansive grasslands did not exist within SLW at this time. Small amounts of cupressaceous pollen (ca. 1.5%) appear for the first time in the zone (Figures 3 and 4). It has been previously suggested that this increase may relate to the expansion of Juniperus, possibly the xerophyte Juniperus scopulorum (Allen, 1995; Hebda, 1995), though more recently, Lucas and Lacourse (2013) suggest that it may represent T. plicata. Either way, the low percentage values suggest that the source trees were relatively minor constituents of the forest. Selaginella prevailed on rocky cliff-faces and slopes at this time, further signalling an open environment. Overall, the data indicate that a fire-maintained open-structure ecosystem emerged during the early Holocene interval. Several other sites within the study region record similar vegetation assemblages (Brown et al., 2008; Cwynar, 1987; Lucas and Lacourse, 2013; Mathewes, 1973), implying that this ecosystem was extra-locally extensive. However, it is difficult to discern exactly how the vegetation and fire regime were expressed across the landscape. While it is possible that open-canopy Pseudotsuga-dominated forest with grassy understory prevailed widely, it is also possible that an aspect-differentiated mosaic developed, with open forest and small grassy areas persisting on southern and western exposures and more closed forest on northern and eastern exposures.
The increases in Pseudotsuga, Poaceae and possibly Juniperus pollen together with Pteridium spores (BL-3; Figure 3) coincide temporally with elevated background charcoal and a moderate, though sustained, magnetic susceptibility signal, suggesting that surface fires of low-moderate severity may have prevailed at this time. Further, the peak component reveals that high-severity crown fires events occurred locally at a multicentennial scale (200–600 year range; Figure 3). In the scenario of widespread open Douglas-fir forest, the fire regime likely consisted of frequent understory burns together with infrequent canopy fires. In contrast, in the scenario linking the distribution of vegetation to aspect, it is possible that frequent lower-severity surface fires prevailed on hot and dry exposures and that crown fires burned less frequently elsewhere in cooler and/or moister settings. It is further posited that, given the warmer summer temperatures during the early Holocene, the prevailing grasses would have cured relatively early in the season, elevating fuel consumption and potential rate of spread (Taylor et al., 1996). The moderate magnetic susceptibility values apparent at this time are likely a result of both the open, dry character of the landscape as well as fire disturbance, with sediment delivery to the basin facilitated by aeolian transport and/or post-fire overland flow.
Other nearby sites similarly exhibit early Holocene increases in fire-adapted taxa (i.e. Pseudotsuga, Alnus, Epilobium, Pteridium), charcoal flux, fire frequency and/or pollen richness as a measure of vegetation diversity potentially reflecting the disturbance regime (Allen, 1995; Brown and Hebda, 2002b; Brown et al., 2008; Leopold et al., 2016; Lucas and Lacourse, 2013; Pellatt et al., 2001). These changes likely indicate that increased fire disturbance typified the local environment at this time. Similar evidence is likewise noted from more distant sites, including northern Vancouver Island (Hebda, 1982), east on the mainland (Hallett et al., 2003; Mathewes, 1973), west on the outer coast (Brown and Hebda, 2002b) and south on the Olympic Mountains (Gavin et al., 2001) and Cascade Range (Cwynar, 1987), Washington State, the US. These spatial patterns likely reflect an overall regional increase in biomass burning during the early Holocene interval.
Human and climate drivers of fire disturbance in the mid- and late Holocene
In the mid Holocene, climate cooled and moistened (COHMAP, 1988; Hebda, 1995; Rosenberg et al., 2004), with changes in orbital geometry leading to reduced seasonality. The east–west precipitation gradient that characterizes south Vancouver Island today was established at this time (Brown et al., 2006). As climate changed, so too did streamflow, with the low flow season shortening (June–September) and discharge increasing (Figure 5). By inference, this change could reflect a shortening of the fire season. The charcoal peak component suggests a continuation of periodic crown fire disturbance (BL-4; 6180–8020 cal yr BP; Figure 3). However, the slight decrease in background charcoal suggests a decrease in high-frequency surface fires. In response to changes in climate and fire regime, fire-adapted species decreased in abundance, whereas fire-sensitive species such as T. heterophylla and T. plicata increased, with the latter documented through an increase in cupressaceous pollen (Figures 3 and 4). The widespread increase in these shade-tolerant species coupled with declines in indicators of open conditions such as Poaceae and to a lesser extent Rosaceae indicates that the forest canopy was closing. The decrease in magnetic susceptibility is likely related to both the overall closure of canopy and reduction of disturbance, consistent with other local and regional sites (Allen, 1995; Brown and Hebda, 2002b; Brown et al., 2008; Cwynar, 1987; Gavin et al., 2001; Hallett et al., 2003; Hebda, 1982; Leopold et al., 2016; Lucas and Lacourse, 2013; Mathewes, 1973; Pellatt et al., 2001). The peak charcoal component suggests that fires continued to burn locally, perhaps partially in response to the arrival and expansion of fire adapted Quercus garryana, a xerophyte, within or near the watershed. At Begbie Lake, Q. garryana pollen values are highest at this time, consistent with other sites in the region that document an oak maximum interval at about 7000 cal yr BP (Allen, 1995; Lucas and Lacourse, 2013; Pellatt et al., 2001). Thus, even though regional climate was moistening, fire-adapted Q. garryana expanded, potentially in response to decreasing seasonality (Pellatt et al., 2001). While more extensively distributed than at present, Q. garryana would have prevailed in dry microsites, including on rocky bluffs, forming savanna-type environments that were maintained by frequent surface fire.
By about 6180 years ago (BL-5; 2500-6180 cal yr BP), cooling and moistening facilitated greater streamflow. The flora was similar to the modern assemblage (Figures 3 and 4), though plant abundances were different. Notably, T. plicata and Alnus were more abundant compared with present, as were rosaceous types. P. contorta and Quercus decreased in abundance. Salix continued to prevail around the basin. The intermittent Lysichiton signal likely reveals the development of a forest swamp ecosystem. In contrast, pollen from T. mertensiana, which likewise occurs intermittently throughout the zone, was likely produced by trees growing at higher elevation and transported down-slope, signalling that cool, moist conditions characterized higher elevations, consistent with other studies examining high-elevation communities (Brown and Hebda, 2003). The increase in Cyperaceae pollen may reflect the development of riparian sedges in shallow and seasonally inundated shoreline.
In contrast to the vegetation, the fire regime between ca. 6180 and 2500 cal yr BP differs notably from both the preceding and following intervals of time (Figure 3). The marked increase in fire frequency reflects increased local biomass burning. A corresponding increase in magnetic susceptibility potentially signifies an increase in burn severity or greater amounts of postfire sediment transport under wetter conditions. Further, the increase in background charcoal could reflect additional surface fires. Given that climate was moistening and cooling, it is difficult to determine what was driving the increase in burning at this time, though it could be related to increased climate variability, human activity, or both (Brown and Hebda, 2002a; Hebda and Mathewes, 1984; Sugimura et al., 2008; Walsh et al., 2015). Regarding the former, while the long-term trend in the mid Holocene was cooling and moistening, the frequency of shorter-term climatic variability events such as El Niño/Southern Oscillation (ENSO) increased between 7000 and 1200 cal yr BP, decreasing thereafter (Moy et al., 2002). In the Pacific Northwest region of North America, El Niño and positive Pacific Decadal Oscillation winters and springs are often warmer and drier compared with long-term averages (Moore et al., 2010), contributing to less snow pack, decreased summer moisture availability and potentially more fire (Haughian et al., 2012; Meyn et al., 2010), possibly of high severity. Regarding human activity, Q. garryana pollen occurs intermittently throughout this interval, in low values. Other sites similarly note low or decreasing Q. garryana pollen values between about 6250 and 2000 cal yr BP (Allen, 1995; Lucas and Lacourse, 2013; Pellatt et al., 2001). In contrast, Pellatt et al. (2001) document a slight increase in oak pollen accumulation between roughly 4000–2500 cal yr BP. It is posited that humans may have been using surface fire to maintain the sustenance-contributing oak ecosystems, promoting an eco-cultural landscape (McCune et al., 2013; Pellatt and Gedalof, 2014; Weiser and Lepofsky, 2009) that facilitated consumption of, for example, acorns and Camassia quamash and Camassia leichtlinii roots (Turner, 1995, 1999). Further, understory vegetation may have been deliberately burned to increase berry yield and browse for deer. Indeed, the discovery of a buried hearth on nearby Orcas Island, US, illustrates a domestic use of fire by humans in the post-Mazama interval (Kenady et al., 2011). Thus, it is suggested that both climate and people were regulating fire disturbance during the 6180–2500 cal yr BP interval.
Establishment of modern conditions
Modern climatic conditions and stream discharge rates were established in the late Holocene (BL-6a; 260–2500 cal yr BP; Figures 3 and 5) and the forests became dominated by P. menziesii together with shade-tolerant, fire-sensitive T. heterophylla and T. plicata. Abies, likely A. grandis, a moderately shade-tolerant species, also increased in abundance, whereas Q. garryana and Alnus decreased. Other sites within the region show comparable trends (Allen, 1995; Brown and Hebda, 2002b; Leopold et al., 2016; Lucas and Lacourse, 2013). Under the canopy, damp swamp ecosystems expanded, as evidenced by the increase in Lysichiton. Commensurate with these changes in vegetation, the fire regime likewise changed in the late Holocene, with fire disturbance becoming less frequent, except in extant Q. garryana meadows, where aboriginal land-management practices involved fire (Pellatt and Gedalof, 2014). The peak component indicates that the FRIs averaged 270 years (Figure 3). This estimate is in general agreement with other estimates of FRIs (ca. 200 years) for dry Pseudotsuga-dominated (CDF and CWHxm) coastal forest (Parminter, 1990; Wong et al., 2004). Regional archaeological evidence indicates that First Nations people were constructing large wooden structures by 3000 cal yr BP and that full-scale working of massive timber using a variety of tools began at 2500 cal yr BP (Hebda and Mathewes, 1984). It is worth noting that the timing of such wood working projects coincides with the end of the 6180–2500 cal yr BP interval of elevated burning. One tantalizing hypothesis would be that the change in human activity somehow relates to the reduction in burning, suggesting that human modification of the local fire environment was somewhat more important compared with climatic regulation. Archaeological evidence clearly reveals that humans actively hunted and camped around SLR over at least the last two millennia (Vincent et al., 2002).
Reconstruction of the recent spatial disturbance history of SLW (Smiley et al., 2016) shows that starting in the AD 1800s to 1940, large areas within the watershed were burned by wildfire. Post-harvest broadcast burns occurred from AD 1920s to the 1970s, whereas more extensive harvesting and pile burning occurred from AD 1970s to the late 1990s, after which harvesting ceased. Forest harvest and pile burning were also associated with reservoir creation and expansion in AD 1915, 1970, 1980 and 2002. These events represent important contributing factors affecting charcoal deposition and vegetation during this period. Indeed, at the very top of the core (BL-6b; –64 to 260 cal yr BP), there is an increase in both charcoal and Alnus, representing European colonization and forest clearance. Forest clearance enabled the pollen signal from Rosaceae, Poaceae and Asteraceae to emerge. In the most recent decades, emphasis has been on fire suppression, though this signal is not apparent in the record.
Fire risk and forest management
Given modern anthropogenic changes in the SLW coupled with climate projections, fire represents a potential risk to water supply, with large, severe fires posing the greatest threat. For example, fire-induced loss of vegetation cover could reduce precipitation interception, evapotranspiration and infiltration rates (Shakesby and Doerr, 2006). The resultant increase in overland flow could enhance erosion and transport of nutrients and other elements to the reservoir (Smith et al., 2011), potentially altering aesthetic characteristics of the water and interfering with the water disinfection process. Enhanced nutrient and sediment input could trigger (potentially toxic) algal blooms or increase total suspended solids, respectively. Post-bloom algal decomposition could increase levels of total coliforms (CRD, personal communication), potentially exceeding water quality guidelines and depleting dissolved oxygen. To mitigate the risk of wildfire within the watershed, the CRD is employing a multi-pronged strategy, with focus on wildfire prevention, response, fuel management and pre-planning for post-wildfire rehabilitation (CRD, 2015). The former involves law enforcement, public awareness, education, fuel hazard mitigation, patrolling for detection and a closed watershed policy to limit the risk of human ignition. To limit fire impact on water, the CRD has developed plans to rapidly respond to any fire within its jurisdiction, with the goal of quickly suppressing any fire and limiting total area burned. Regarding fuel management, the CRD is creating fire breaks to prevent fire spread within SLW and has conducted forest-fuel reduction around key facilities using the FireSmart methodology (Partners in Protection, 2003). To reduce the potential for post-wildfire impacts, the CRD is modelling sediment and debris flow potential, identifying risk mitigation and rehabilitation strategies, techniques and locations where they would be used on the landscape.
Considering the future, marked changes in climate and vegetation are regionally anticipated. Summer and winter temperatures are projected to increase, with increased summer drying and more frequent extreme precipitation events (Vines et al., 2017). Ecological climate niche simulations suggest that the CDF zone may expand westward on south Vancouver Island (Hebda, 1997b), possibly into or around SLW (Wang et al., 2012). In addition, Pellatt et al. (2012) indicate that habitat suitable for Q. garryana could also increase on southeastern Vancouver Island, signalling regional forest opening. In addition to climate and ecological change, the population of greater Victoria is likely to grow (CRD, 2014), potentially increasing the risk of human-caused ignitions. The abovementioned changes will all variously contribute to altering the fire regime, with several studies suggesting that fire disturbance may increase in southern BC in the coming decades (Flannigan et al., 2005; Haughian et al., 2012; van der Kamp et al., 2013; Wang et al., 2016). While not an exact analogue for the future due to, for example, differences in climate forcing mechanisms, introduction of invasive species and land fragmentation (Brown et al., 2017; Gavin et al., 2007) the paleoenvironmental record from Begbie Lake nevertheless provides valuable insight into the interaction among climate, vegetation, fire and people through time in SLW. Such insights augment both present-day monitoring efforts, management practices and model simulations, generating a range of perspectives that managers can consider when developing climate adaptation strategies and fire prevention policies.
Regarding management of SLW, the Begbie Lake paleoenvironmental record, combined with insights from paleoclimate models and forecast simulations, suggest that the vegetation and fire regime will change in the future for the following reasons: (1) Forest canopies may become more open and grasses may expand, creating a mixed-severity fire regime of crown and surface fires; (2) the fire season will likely lengthen; (3) drought-tolerant and fire-adapted species such as P. menziesii may increase in abundance in response to regional drying. Likewise, while not common in the watershed today, fire-promoting Pteridum could also increase in abundance in response to increased fire disturbance and associated canopy openings; (4) species that occur within the watershed today like T. heterophylla and T. plicata will likely become increasingly stressed in the future. Mortality will add fuel to the forest floor and possibly create ladder fuels if dead trees and branches are not removed, increasing fire risk; (5) the introduction of exotics, such as the shrub C. scoparius, presents an unprecedented fire risk; (6) People have historically burned in the watershed, as suggested by the mid-Holocene increase in fire contemporaneous with the presence of intermittent Quercus pollen. While SLW is protected, future community growth near the watershed will increase the risk of human ignition.
Finally, species that are well-suited to warmer and drier conditions and which are adapted to fire should be promoted within the watershed, potentially reducing fire impacts. For example, post-fire survival of P. menziesii, a fire resistor due to its thick bark, would mitigate fire impacts by surviving burns and increasing post-fire precipitation interception and evapotranspiration (Shakesby and Doerr, 2006). Also, while the Holocene record of G. shallon (salal; ericaceous-type pollen) is poorly resolved due to low representation, it is at present the dominant understory shrub within SLW. G. shallon has a high moisture content, low amounts of sap and resin and produces little dead wood (Detweiler and Fitzgerald, 2006). Thus, its presence in the watershed may help limit fire spread, though additional work is required to better understand the fire fuel potential of salal, especially during periods of extended drought. It should also be noted that potential lengthening of the fire season, as suggested for the past and predicted for the future, implies that less time may be available to mitigate the impacts of any undesired fires within the watershed prior to onset of heavy winter rains, necessitating preparation. Educational campaigns utilizing data from studies such as this, should likewise be developed so that citizens become aware of fire regime change and impacts of fire, as well as benefits of controlled fuel treatment.
Conclusion
Collecting a sediment core from a lake within a managed municipal water supply watershed added an applied component to the investigation. The study examined how regional climate as well as watershed vegetation and fire disturbance changed through time in response to various drivers including climate change and variability as well as human actions such as First Nations burning practices and European settlement. Investigations such as this provide watershed managers with information about climate-vegetation-human-fire interactions and historical fire disturbance within the watershed that would otherwise be unavailable.
Relating paleoecological data to resource management is often challenging. In this instance, insights from the warm, dry early Holocene provide context for modern and future conditions induced by climate change. The Begbie Lake record reveals that fire-adapted ecosystems prevailed in and around the watershed during the early Holocene, commensurate with a longer fire season and more regional fire disturbance. Given that both model simulations and paleoecological data indicate that watershed vegetation and landscape processes are sensitive to changes in climate, managers need to prepare for the upcoming transformation.
Footnotes
Acknowledgements
The authors wish to thank the Capital Region District for providing funding for this work through Service Agreement PFC-CRDIWS-4. Additional funding was provided by Natural Resources Canada. The authors wish to extent their gratitude to Joel Ussery and Rob Walker of CRD for their support, participation and various other contributions. We also extend our gratitude to Bryon Smiley and Kangakola Omendja of Natural Resources Canada for assistance with GIS mapping. Brian Wiens of Natural Resources Canada and Richard Hebda commented on an earlier version of the article. Finally, we wish to extend our gratitude to the reviewers who provided constructive comments that improved the manuscript.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
