Burn Severity Does Not Significantly Alter Pollen Abundance Across a Burn Matrix Four Years Post Wildfire in Sub-Boreal Forests of British Columbia, Canada

Abstract

Wildfires have had measurable impacts on pollen dispersal in some areas; both facilitation and potential barriers to pollen movement have been reported. These dispersal dynamics in turn affect population genetics and reestablishment of seed-producing plants, at times significantly impacting the successional trajectory of the area in question. However, research on post-fire pollen distribution and occurrence is lacking for the boreal and sub-boreal forests of western Canada, and many communities that have been heavily impacted by wildfire remain concerned about the future forest landscape of these areas. We analyzed post-fire pollen samples from unburned and severely burned sub-boreal spruce stands in north-central British Columbia four years after a major wildfire. We used pollen traps to measure the occurrence and abundance of pollen types from four important plant families: Asteraceae, Ericaceae, Onagraceae, and Pinaceae families, to address specific concerns of the First Nation communities with territories overlapping the Shovel Lake wildfire burned area. Pinaceae pollen was found across all traps and was observed as the most dominant pollen type at all study sites, while pollen belonging to other families was found less frequently. No significant differences in pollen occurrence or abundance were found between burn severities, despite differences in the plant communities; however, plant and pollen abundance were found to be positively correlated to one another. These results may indicate that, as previously noted in other conifer-dominated forests, openings of the forest landscape by wildfire may facilitate rather than hinder pollen movements. Understory species should be studied in more detail as the effect of wildfire on pollen transport may vary between taxa and pollination syndromes.

1. Introduction

1.1. Importance of Fire in Sub-Boreal Ecosystems

Fire is a dominant cause of disturbance in the sub-boreal spruce forests of central interior British Columbia, with stand-replacing fires typically occurring every 230 years or less depending on location [1,2]. These natural disturbances are known to maintain a diversity of stand ages and structures across the landscape and therefore promote biodiversity [1]. Early successional stands are generally dominated by lodgepole pine (Pinus contorta Dougl.) seedlings, which are often reestablished from serotinous cones; shrubs and herbaceous species, which survive via underground structures such as rhizomes and through soil seed banks, are also common [1,3]. In general, there is a decrease in herbaceous plant cover and understory species richness, and an increase in structural complexity as succession progresses [1]. Fires are a natural and important component of sub-boreal forest ecosystems, but the frequency, severity, and size of forest fires in British Columbia (BC) are currently increasing due to several factors, including the changing climate, buildup of coarse woody debris from trees killed by the mountain pine beetle (Dendroctonus ponderosae Hopkins) epidemic, and past fire suppression [4]. Therefore, the rate of plant community recovery between fires is of increasing relevance to sub-boreal forest ecosystems, and the people who rely on them.

Unburned forest patches are critical for forest regeneration and influencing the ecological succession of surrounding severely burned areas, as they provide sources of seeds and other propagules that disperse into burned areas [5]. Reproductive success of both pioneer and mature forest species is crucial for forest regeneration. Many dominant plant species in sub-boreal ecosystems, such as conifers, rely on seeds for reproduction and dispersal; other species that can also reproduce asexually can typically disperse farther using seeds [6], making sexual reproduction of particularly high importance for repopulation of burned areas. A critical step in successful seed production is successful pollination.

1.2. Importance of Pollen Dispersal for Plant Reproduction

Although some species of seed plants are capable of self-pollination (autogamy), many have mechanisms that prevent self-pollination to maintain beneficial genetic diversity and prevent inbreeding depression in populations [7,8]. In terrestrial sub-boreal ecosystems, most pollen is transported by wind (anemophily) or insect mutualists (entomophily) [9,10].

Many dominant tree species of BC’s sub-boreal forests, including interior hybrid spruce (Picea engelmannii × glauca [Moench.] Voss), lodgepole pine, subalpine fir (Abies lasiocarpa (Hook.) Nutt.), and paper birch (Betula papyrifera Marshall.) are considered anemophilous [11,12]. Pollen grains of coniferous tree species possess paired air sacs, or sacci, which appear to augment the pollen’s aerodynamics such that it can be carried over great distances by wind [13]. In contrast the understory plants of boreal and sub-boreal forests are typically entomophilous outcrossing species, relying on bumblebees, flies, and other insects for pollen dispersal [10]. The exine’s surface features, along with overall size and proportions of the pollen grain, are the main characteristics used for taxonomic identification of pollen [14,15].

Many factors influence the dispersal of pollen grains across the landscape. The pollination adaptations of a species may influence the physical structure of pollen grains [16,17], but they also influence the number of pollen grains produced and how far they can disperse from the plant. For example, the number of pollen grains produced by anemophilous outcrossing species can be orders of magnitude greater than those produced by autogamous species [14]. The number of pollen grains produced by zoophilous species varies greatly, from very few to amounts comparable with anemophilous species, especially in zoophilous species that attract pollinators by providing excess pollen as a food source, and those that are considered facultatively anemophilous [14]. Pollen of zoophilous species is typically transported over much shorter, targeted distances as it is influenced by the complex behavior of pollinator species [18,19] compared to that of anemophilous species [9,14]. Anemophilous pollen is generally broadcasted into the surrounding environment and final dispersal patterns are strongly influenced by wind direction and velocity, topography, and physical attributes of the surrounding plant community such as height and stand density [14,20,21]. Due to the number of grains produced, directionality of transport, and average dispersal distances, pollen samples from any random location in the sub-boreal landscape will typically contain a higher abundance of anemophilous and facultatively anemophilous species compared to zoophilous and autogamous species unless the deposits are collected from locations very near zoophilous or autogamous plants [14].

Previous studies, both modern and paleopalynological, have found dramatic changes in pollen dispersal and species composition of forest habitats after wildfires [18,20,22,23]. In a Mediterranean pine forest, it was demonstrated that pine pollen dispersal was possible over longer distances after fire, likely due to more open habitat structure which increased airflow and reduced the number of physical obstacles to anemophilous pollen dispersal [20]. It is also important to consider the successional timeline and how long these dispersal effects may last. Furthermore, the effects of fire on pollen dispersal may be more complex and varied for insect-pollinated plants, with both negative and positive responses in abundance and activity reported for different insect groups [18,19]. Therefore, the effects of fire on insect-facilitated pollen dispersal are likely to vary based on the unique pollinator interactions of each plant species. The pollination syndrome of a plant species may be a determining factor as to whether recently burned areas act as corridors or barriers for pollen transfer.

1.3. The Shovel Lake Wildfire Area

The Shovel Lake wildfire occurred north of Fraser Lake, BC, within the territories of the Stellat’en and Nadleh Whut’en Nations, among others, between July and September of 2018. The fire covered an area of approximately 92,000 hectares [24]. The burn perimeter included many lakes and recreation sites, as well as important cultural and harvesting areas for these Nations. The forest within the fire perimeter mainly consisted of lodgepole pine- and interior hybrid spruce-dominated stands, ranging in age from recent clearcuts (stands harvested for timber acquisition) to forests over 100 years old, with high levels of past and present forest harvesting activity [25]. The burned area can be generally described as a matrix landscape, with burn severity pockets minimally, moderately, and severely burned. The combined disturbances of logging and fire have had strong impacts on the forest ecosystems, and the local people who rely on them for cultural and economic value; therefore, there is strong interest in understanding ecosystem recovery in this area. The people of this area voiced concerns over the potential for the vegetation to return to its natural state, given the anthropogenic disturbance that had occurred in the territories, the subsequent mountain pine beetle epidemic, followed by the wildfire disturbance [26]. Specifically, the First Nation communities in the Shovel Lake fire area expressed interest in the health and recovery of heart-leaved arnica (Arnica cordifolia Hook., family Asteraceae), fireweed (Chamaenerion angustifolium L., family Onagraceae), and several species in family Ericaceae including black huckleberry (Vaccinium membranaceum Douglas.), Labrador tea (Rhododendron groenlandicum Oeder.), and dwarf blueberry (Vaccinium caespitosum (A.Gray.) Mart.). The fire regimes and successional vegetation patterns of these forests have been intensively studied in the past, from perspectives of both ecological interest and forest management [1,3], but little to no research exists on how modern pollen dispersal in sub-boreal forests is affected by fire.

1.4. Research Objectives

Many of the previous studies on changes in pollen dispersal after fire have focused on individual species [20], shifts in plant community composition [23], or the activity level of pollinators rather than the physical distribution of pollen [18,19]. In this study we aim to determine pollen dispersal across a post-burn fire matrix for the plant communities in the Shovel Lake fire area of north-central BC, including both wind- and insect-pollinated taxa. We were specifically interested in pollen of families Asteraceae, Onagraceae, and Ericaceae, as these were mentioned as focal species for the communities living in the fire area, and pollen of family Pinaceae, as the dominant conifer regenerating in the burn matrix. Many previous studies [18,19,20] were conducted in Mediterranean forests, and there is a lack of available information on changes in pollen dispersal after fire for boreal and sub-boreal forests. The present study contributes to a deeper understanding of post-wildfire pollen distribution and abundance and lends insight to whether reproductive cycles in specific species are helped or hindered by fire, for conservation and forest management purposes.

The main objectives of the study were to determine the following:

• Whether the quantity and variability of pollen collected, summed from all species, differed between severely burned and unburned forest patches within the fire matrix;
• How the taxonomic composition of pollen samples varied, in terms of relative abundance and frequency of occurrence, between severely burned and unburned forest patches with focus on the four families of interest;
• Whether pollen composition (both frequency of occurrence and relative abundance) and plant community relative abundance follow the same patterns across space and burn severities, with focus on the four families of interest.

Based on these objectives, we predicted that pollen deposits in severely burned forest patches would be greater and have lower variance between pollen traps compared to unburned forest stands. Severely burned patches were also predicted to have a higher relative abundance and frequency of the families of interest due to the ability of pollen to be carried farther over space (limited spatial barriers). We further predicted that pollen of some forest understory species would be represented in pollen traps from unburned stands, but absent in traps from severely burned stands, corresponding to a greater diversity of plants at unburned sites.

2. Materials and Methods

2.1. Study Sites

We established plots in three separate locations within the Shovel Lake fire perimeter north of Fraser Lake, BC (Figure 1). We selected sites within the same biogeoclimatic ecosystem classifications to reduce ecosystem variation between sites. We also looked for locations that would accommodate a paired design with both an unburned and severely burned sampling plot at each location. Burn severity was assessed using satellite imagery from before and after the fire and later confirmed with visual on-site assessment using composite burn indices (CBIs) [25,27]. Severely burned plots had over 90% mortality of mature trees with substantial consumption of coarse woody debris, litter, and organic soil horizons, and CBIs in the high severity range (2.5–3.0). Plots designated as unburned had no visible evidence of recent fires. Severely burned and unburned plots were established according to the organic nature of burn matrix, often following topography, so there was no consistent distance set between each pair. At site 1 (Angly Lake), the severely burned stand was on the crest of a low hill bordering on a wetland environment, and some lower areas of the plot had high soil water content; at site 2 (Trout FSR), the severely burned plot bordered on a cleared area, had a lower slope position, and had lower soil moisture; the third site (Top Lake) was located in a gully running down from a logging road, had high soil moisture, and bordered on a small lake. Distance from the center of each plot to the nearest road and forest edge (clearing, lake, or wetland) was measured using the most recent Google Earth satellite imagery, from 2015–2022 [28] (

Table 1. Landscape characteristics of study plots north of Fraser Lake, British Columbia, Canada within the Shovel Lake wildfire perimeter. The types of forest openings that were considered in the “distance to other opening” category included clearing types other than roads, such as lakes and wetlands.

Plot	Distance to Road (m)	Distance to Forest Edge (m)	Distance to Other Opening (m)	Approximate Slope Position	Relative Soil Moisture
U1	40	453	91	Backslope (mid-slope)	Moderate
S1	47	175	54	Crest	Moderate-high
U2	81	183	570	Shoulder (upper-slope)	High
S2	21	27	27	Foot/toe slope	Low
U3	34	75	29	Flat at toe of slope	Moderate-high
S3	118	45	106	Backslope, in gully	High

). Prior to the wildfire in 2018, all stands consisted of mature sub-boreal spruce forest dominated by lodgepole pine and interior hybrid spruce (stand age ≥ 80 yrs, based on core samples). Plots were named using a two-character system, with a letter denoting burn severity (U for unburned, S for severely burned) followed by the site number (1, 2, or 3).

2.2. Pollen Sampling

Within each study plot we established two 50 m long transects, crossing at a 90° angle at their centers to cover as much of the plot as possible. We placed nine pollen traps along the transects at each plot: one at the intersection of the two transects, and every five meters along the transects radiating out from this point (Figure 2).

The pollen trap design was based on the modified Oldfield trap [29] (Figure 3). Each trap consisted of a plastic container with an 8.3 cm diameter upper opening, over which 1 × 2 mm fiberglass window screen was secured with 24-gauge galvanized steel wire to prevent the entry of large contaminants and hold the trapping material in place. A 16 mm diameter drainage hole was drilled in the bottom of each container, over which a cellulose filter circle with 8 μm pore size was secured using a silicone all-purpose sealant. A second piece of window screen was glued in place over the bottom of the drainage hole to prevent damage to the filter paper (Figure 3). The trap was filled with loose rayon fiber-trapping material [29,30]. Each trap was positioned with the opening as level as possible with the forest floor surface to prevent movement of the traps during the sampling period. We set the traps out in the study plots on 9 June 2022, and left them for 5 weeks (35 days) to capture a variety of pollen types.

Our study region has a relatively short growing season, only about 100 days, compared to other parts of British Columbia [31]. The timing for pollen trapping was selected as it spans the main period of anthesis for most forest species in this region. Although anthesis varies slightly year to year, we made careful observations of plants at our study sites in 2022, to ensure we were capturing the flowering time for the targeted families. After collection, the traps were stored at 3 °C until processing. We also collected floral samples from some plant species that were available within the Shovel Lake fire area, and in flower at the time of data collection, to use as references for pollen identification.

2.3. Pollen Sample Processing

Pollen from available reference flowers and from pollen traps was processed at the University of Northern British Columbia’s Plant Lab. Reference pollen grains were extracted from the anthers of dried floral specimens and rehydrated in glycerine for at least 10 min before photographing using Nikon^® NIS Elements Basic Research software (v.5.10.01 64-bit). Some pollen traps contained large amounts of water at the time of collection, which we removed using vacuum filtration with a sink aspirator and 8 μm pore-size cellulose filters. Both the filters and rayon fibers were kiln-dried at 50 °C for 24 h before undergoing a standard acetolysis procedure [15]. A standard mixture of 9:1 volumetric ratio of 97% acetic anhydride to 95% sulfuric acid was prepared, and 40 mL of the mixture was added to each sample. Suspended samples were heated to 80° C in a hot water bath for 6 min to dissolve the rayon fibers, filter paper, and cellulose-based contaminants [15]. The samples were then divided between three 15 mL centrifuge tubes and centrifuged at 2900 rpm for two minutes to separate pollen grains from the solution, after which the acetolysis mixture was decanted and discarded. Pollen pellets were rinsed in 5 mL of glacial acetic acid to remove residual acetolysis mixture, centrifuged and decanted, and rinsed three times with 10 mL of deionized water following the same centrifuging and decanting procedure [15]. Paper coffee filters were used to cover the centrifuge tubes to prevent contamination, and samples were dried in a 50 °C kiln for approximately 48 h. Samples were then weighed, re-suspended in 5 mL of glycerin (approximately 1.67 mL per centrifuge tube), and re-combined for storage and mounting on microscope slides. The acetolysis processing method only dissolved organic materials, and mineral soil particles remained in some samples from severely burned plots after processing, as exposed mineral soils in these plots were more easily blown into the traps. In some cases, these soil particles obscured some pollen grains present, making counting and identification more difficult.

One slide was prepared per pollen trap using a subsample of the trap contents and analyzed using light microscopy (NIS Elements BR software (v.5.10.01 64-bit) and Nikon^® Plan Fluor 10x0.30 DIC L/N1 and 20× objective lenses) to determine the taxa of pollen present and their relative densities. Due to the limited references available on pollen identification for northwestern Canada, the taxonomic groups identified were restricted to the families of interest: the Asteraceae, Ericaceae, Onagraceae, and Pinaceae, with some estimation of an additional family, Rosaceae, due to its prominence in the plant community. All other pollen grains were only counted as part of the total pollen abundance. We used dichotomous keys [14,32], reference samples, and visual resources from the Global Pollen Project reference collection database [33] to determine the defining visual characteristics of these pollen types for identification, presented in

Table 2. Pollen characteristics for the four families of interest. Pollen was positively identified in the pollen traps collected from the Shovel Lake wildfire in northern British Columbia, Canada. Characteristics listed were used for microscopic identification of pollen. In the Rosaceae family, due to the diversity of pollen types, we provide information for the genus Rubus, as this was one of the most prevalent genera represented from Rosaceae in the study area.

Family	Diameter	Shape	Ornamentation	Aperture Type
Pinaceae	70–100 µm	Saccate	Any	NA
Asteraceae	20–30 µm	Oblate to suboblate or peroblate	Echinate (or echinolophate)	Tricolporate
Onagraceae	55–65 µm	Spherical to sub-triangular	Psilate	Tetraporate
Ericaceae	29–45.3 µm (entire tetrad)	Spherical, arranged in permanent tetrads	Any (Ericaceae is the only plant family with permanent pollen tetrads that is native to the region)	Any
Rosaceae	In Rubus: Average length of polar axis is 27.24 ± 7.51 μm and equatorial axis is 20.47 ± 4.58 μm [35] (Xiong et al. 2019)	Prolate to oblate ellipsoidal, most genera have subspherical pollen [36] (Hebda and Chinnapa 1994)	In Rubus: Perforations are large and often extending onto tectal ridges; smooth or verrucate sculpturing with pores [36] (Hebda and Chinnapa 1994)	Tricolporate (rarely tricolpate) [36] (Hebda and Chinnapa 1994)

. The Rosaceae family contains diverse forms of pollen, boundaries of which are not well delimited even within genera [34], and no key for the study region was available for certain identification of pollen from this family. Pictures from the Global Pollen Project database, of pollen from specific Rosaceae species known to be highly prevalent across the study sites, were used to identify these pollen groups to the best estimate possible. Only pollen grains in which all defining features were visible at 200× total magnification were counted as positive identifications.

To determine the presence of pollen grains from different families in each trap, we visually assessed the entire slide area at 100× magnification, and when unique pollen grains were located, we captured images of these at 200× magnification with Nikon’s NIS Elements Basic Research software (v.5.10.01 64-bit). Following this, 36 additional images were systematically acquired at 200× magnification, spaced evenly across the slide, and comprising a total area of 0.09 cm². These were considered as a representative subsample of the pollen on the slide and were used for absolute pollen counts and obtained a relative abundance of identified and unidentified pollen grains in the sample.

2.4. Plant Community Data Collection

To measure plant community richness and abundance, we established a 20 × 20 m square plot next to each pollen sampling plot and used a transect and quadrat method of systematic random sampling. The two transects were placed across the plot perpendicular to the topographic slope direction, and species presence and percent cover were recorded within a 0.5 × 1 m quadrat every two meters along each transect, for a total of 20 vegetation quadrats measured per plot.

2.5. Statistical Analysis

We were confident in identifying the Asteraceae, Ericaceae, Onagraceae, and Pinaceae pollen types, so counts and relative abundance calculations were performed; remaining unidentified pollen were grouped as “other pollen” for analysis. We also provided an estimate of overall Rosaceae pollen within the “other pollen” category for our descriptive analysis, as this plant family was found abundantly within the sample area. We did not fully quantify Rosaceae as we were less confident in absolute counts of Rosaceae pollen due to the lack of dichotomous key available for the region. The relative abundance of pollen in each trap (sample) was calculated by family as the total number of pollen grains divided by the total number of pollen grains observed of all types in the microscope images. For Ericaceae pollen, we counted each observed tetrad as a single dispersal unit due to the presumed permanence of the tetrad structure [32]. The arithmetic mean density (x̄) and standard deviation (σ) were then calculated for pollen abundance by family for each site. Presence or absence and total percent cover of the plants from the dominant pollen families observed growing in the plots were calculated for each quadrat, after which we calculated the relative abundance and the frequency of occurrence across each sampling area. All data were assessed for normality by treatment, species, and site using the Shapiro–Wilk test. Due to the non-normal distribution of the majority of the data, we compared the pollen counts and relative abundance of plant families by site and by family using a Kruskal–Wallis test and tested the difference between unburned and severely burned treatments using a Mann–Whitney U test. Frequency of pollen occurrence was compiled as count data (number of traps in which a given family’s pollen occurred) at each sampling plot, using a slightly reduced dataset so that the same number of traps (5—center and transect ends) were included per plot. We then used Poisson generalized linear models to test for statistically significant effects of burn severity (unburned/severe) and percent cover of the corresponding plant families on frequency of pollen family occurrences per plot. Pinaceae pollen occurred in every trap across all sampling plots; therefore, occurrence of Pinaceae pollen was not included in the frequency analysis.

The percent cover of pollen-producing plants was compared to the pollen abundance data using correlation analysis, for unburned and severely burned sites, to determine if the existing plant communities proportionally influenced the occurrence of pollen. A list of the genera for each family of interest was curated. Statistical tests were performed using RStudio with R version 4.1.2, dplyr package [37], and readxl package [38], and using IBM SPSS version 29.0.

3. Results

We successfully photographed and analyzed pollen from at least five traps at each site sampled (Figure 4). We obtained pollen from all nine traps at sites S1 and U2, for eight traps at sites U1 and S2, and for five traps at sites U3 and S3. The total abundance of pollen counted across all traps was lower than anticipated. A total of 714 pollen grains were counted across all sites; 53.6% of the pollen represented was from the Pinaceae family and 45.1% was not officially identified, but we estimate that approximately 5% of this unidentified pollen belonged to the Rosaceae family when applying the characteristics from Table 2. The remaining pollen was identified as either Ericaceae (1.0%), Onagraceae (0.15%), or Asteraceae (0.15%). Figure 5 shows the relative abundance of pollen from different families within each treatment group.

A greater number of pollen grains were detected from the severely burned sites compared to the unburned sites (410 vs. 304, respectively); however, these groups of total pollen were not significantly different (Z = −0.343, p = 0.744). Similarly, no significant difference was noted across the relative abundance calculations of vegetation community (plant families of interest with an additional “other plant” category) based on treatment (Z = −0.021, p = 1.00). Pollen and plant relative abundances were not significantly different between site locations (1, 2, or 3; see Figure 1) (pollen; H = 0.671, p = 0.795) (plant; H = 0.180, p = 0.194); however, both pollen and plant relative abundance were significantly different by family (pollen; H = 24.101, p < 0.001), (plant; H = 10.268, p = 0.036).

In all study plots, Pinaceae pollen was the most abundant of the four families quantified, while most of the remaining pollen originated from other taxa outside of the families of interest (labeled as “other” in

Table 3. Average pollen abundance with standard deviation by family. Pollen was collected from the Shovel Lake wildfire burn matrix and identified from families of interest to the surrounding communities in the Fraser Lake area of north-central British Columbia, Canada.

		Plant Relative Abundance (In Treatment)		Pollen Relative Abundance (In Treatment)
Family	Treatment	Mean	Std. Deviation	Mean	Std. Deviation
Asteraceae	Severe	1.985%	1.911%	0.081%	0.141%
Ericaceae	Severe	2.878%	3.919%	0.407%	0.373%
Onagraceae	Severe	9.089%	2.893%	0.081%	0.141%
Other	Severe	11.862%	4.830%	15.935%	7.562%
Pinaceae	Severe	1.177%	1.143%	16.829%	12.577%
Asteraceae	Unburned	0.962%	1.401%	0.000%	0.000%
Ericaceae	Unburned	10.252%	3.367%	0.219%	0.190%
Onagraceae	Unburned	0.216%	0.143%	0.000%	0.000%
Other	Unburned	12.451%	8.087%	13.816%	10.542%
Pinaceae	Unburned	8.068%	7.148%	19.298%	10.492%

). This “other pollen” included family Rosaceae, for which we estimated that within site 1, Rosaceae pollen accounted for approximately 20% of the “other pollen” group in the unburned plot and approximately 8% of the “other pollen” in the severely burned plot. The abundance of pollen measured for Asteraceae, Ericaceae, and Onagraceae was low across all sites.

All families of pollen were found in both unburned and severely burned plots, but at different frequencies. Pinaceae pollen was the only type found in every pollen trap, across all sampling plots.

Table 4. Percentages of traps in which Asteraceae, Ericaceae, and Onagraceae pollen grains were found for each study plot (% of traps) within the Shovel Lake wildfire burn matrix. The wildfire took place in 2018 and pollen was collected in 2022, in north-central British Columbia, Canada.

Site	Burn Severity	Asteraceae Pollen Frequency (% of Traps)	Ericaceae Pollen Frequency (% of Traps)	Onagraceae Pollen Frequency (% of Traps)
1	Unburned	0	25	12.5
	Severe	50	33.3	11.1
2	Unburned	0	33.3	0
	Severe	50	25	25
3	Unburned	40	40	20
	Severe	40	80	80

shows the percentages of traps in which Asteraceae, Ericaceae, and Onagraceae pollen grains were found for each sampling plot. While frequency of pollen occurrence differed slightly between plots (for all but the Pinaceae family), models did not reveal any significant effects of either burn severity or percent cover of the corresponding plant family on pollen occurrence (all p values > 0.09). However, the relative abundance of pollen in each of Asteraceae, Ericaceae, Ongraceae, Pinaceae, Rosaceae (based on site 1 estimate), and “other plant families” was significantly, positively correlated with the relative abundance of plants from these families, across sites and plots (Spearman’s rho = 0.380, p = 0.032).

Even though burn severity did not seem to influence the pollen of the area, the models for vegetation did indicate that frequency of plant occurrence was significantly related to burn severity for the Asteraceae, Ericaceae, and Onagraceae families (p = 0.0071, 0.000675, and 2.1 × 10⁻⁷ respectively). More specifically, the frequency of the Asteraceae and Onagraceae families was negatively associated with unburned treatment (slope estimates = −0.7167 and −1.77495, respectively), while the frequency of the Ericaceae family of plants was positively associated with unburned treatments (slope estimate = 0.8023). The genera found in the sample area are curated in

Table 5. A list of the plant genera belonging to the families of interest (Asteraceae, Ericaceae, Pinaceae, Onagraceae, and Rosaceae) found within the unburned and severely burned sample sites in the Shovel Lake wildfire burn matrix, north-central British Columbia, Canada.

Families of Interest	Genera in Unburned Sample Area	Genera in Severely Burned Sample Area
Asteraceae	Achillea	Anaphalis
	Anaphalis	Arnica
	Antennaria	Hieracium
	Arnica	Petasites
	Aster	Taraxacum
	Petasites
Ericaceae	Arctostaphylos	Orthilia
	Empetrum	Vaccinium
	Gaultheria
	Orthilia
	Rhododendron
	Vaccinium
Onagraceae	Chamaenerion	Chamaenerion
Pinaceae	Abies	Picea
	Picea	Pinus
		Pseudotsuga
Rosaceae	Amelanchier	Rosa
	Fragaria	Rubus
	Rosa	Spirea
	Rubus
	Spirea

for each family of interest.

4. Discussion

The main understory plants selected as the focus of this study are local species belonging to the Asteraceae, Ericaceae, and Onagraceae families. Heart-leaved arnica (Arnica cordifolia Hook.) is a member of the Asteraceae family that is common in low- to mid-elevation forests of northern British Columbia and is particularly abundant after forest fires [39]. Fireweed (Chamaenerion angustifolium L.), a common local member of the Onagraceae family, is often found in open areas such as those created by fire [39]. The Ericaceae family includes many common sub-boreal forest plants, such as black huckleberry (Vaccinium membranaceum Douglas.), Labrador tea (Rhododendron groenlandicum Oeder.), and dwarf blueberry (Vaccinium caespitosum (A.Gray.) Mart.) [39]. These understory species are primarily entomophilous [40,41,42,43,44]. We also quantified pollen for the dominant family making up the forest canopy in this region, Pinaceae, which is represented largely by the species lodgepole pine and hybrid spruce. Pinaceae are anemophilous trees. The high abundance of Pinaceae pollen across all sites and plots was most likely caused by the dominance of pine and spruce in the remaining mature forest patches, and their pollination syndrome. In fact, it was reported that up to 350 million pollen grains may be produced by a 10-year-old Pinus branch system at mid-latitudes [45]. Pollen of anemophilous trees can be lifted into higher air strata by convection currents and turbulence above the forest canopy and occasionally transported over distances of 100 km or more [14]. When warm pockets of rising air collapse, or the air temperature decreases at night, air currents can reverse and deposit large amounts of airborne pollen onto the surfaces below [14]. The lack of canopy cover in recently burned areas would make it easy for these airborne pollen deposits to reach the ground-level traps, leading to high densities of Pinaceae pollen despite the absence of live mature trees. The modified Oldfield trap design also mimics the natural collection of pollen grains by moss and may therefore provide higher representation of anemophilous pollen compared to zoophilous pollen [29].

The results of our analysis did not support our prediction that pollen would be more abundant in severely burned areas. Other studies have varying reports, depending on the ecosystem, dominant plants, and time of measurement. It was reported that the total pollen transported by insect mutualists in burned areas was only 20% of that in unburned areas [18]. However, older studies of pollen samples from soil report that recently burned areas typically had higher numbers of pollen grains per m² compared to intact forest due to dense foliage of forest edges and canopies acting as barriers to incoming pollen [14]. Likely, studies that collected pollen from soil are able to access pollen that has been deposited over long periods of time, contrary to our study which captures only pollen that is moving at the time of collection. However, studies in wildfire-affected areas that retrieve pollen from soil must carefully consider sampling depth; pollen deposits deeper in the soil profile may be more easily accessed in recently burned areas due to the removal of surface vegetation and organic soil horizons, compared to pollen in unburned spaces, thereby biasing sampling numbers. It was reported that Pinus-dominated forests have remained resilient and stable across multiple fires and climate change events through time, indicating that if changes in pollen abundance are realized in the short term, likely these will not persist [23]. Therefore, our finding that pollen abundance was not significantly impacted by fire severity four years post fire indicates that these forests remain resilient, and it can be expected that plant communities that are significantly different between unburned and severely burned sections of the Shovel Lake burn matrix will continue to infiltrate and recolonize the burned areas over time.

The vegetation in unburned plots likely led to more variable deposition of pollen across space, hence the lower frequency of detection. Unburned plots had a greater structural complexity, comparable to what is frequently observed in mature forests of the sub-boreal region, which would have resulted in a higher probability that pollen would be deposited into the traps [1]. Conversely, the consistently open habitat of the severely burned plots may allow more even dispersal of pollen across the landscape. As observed for pine pollen, open spaces such as those created by stand-destroying fires can facilitate greater dispersal of anemophilous pollen [20]. Future studies of modern pollen dispersal post fire could build on our results by incorporating an aspect of the experimental design that may more precisely capture the movement of zoophilous pollen.

We tried to control for environmental variation as much as possible in the placement of our study sites, and our results indicate that there were no significant differences in pollen abundance according to site. It is possible that pollen traps located in areas that bordered on openings such as at site 3, adjacent to a small lake, could have resulted in differences in wind dispersal of pollen compared to the other sites which bordered on roads and forest; however, no significant influence of this was noted in our analysis. This lack of difference further demonstrates that the burn matrix openings facilitated typical movement of pollen across the landscape that was not contrary to what may exist outside burned areas by year 4 post fire.

The differences in plant family occurrence between burn severities can likely be explained by both habitat preferences and lifecycles. Fireweed, along with many members of the Asteraceae family, prefer open habitat and are adapted to quickly establish in disturbed areas through mechanisms such as airborne seeds; conversely, the Ericaceae family contains several slow-growing understory shrubs [39]. The significant positive correlation between pollen and plant abundance for these families indicates that the pollen is being dispersed in proximity to parent plants, which is reasonable.

We acknowledge that the dataset is limited in terms of size, with a total of 714 pollen grains counted. Future studies could examine a greater number of sites over the large spatial areas of wildfire burn matrices in northern BC and incorporate the use of a greater number of pollen traps to further validate our findings. Furthermore, looking at burn sites over multiple years post fire would give an idea as to how the pollen dynamics change over a temporal scale so that we may be able to determine the recovery rate associated with these resilient systems.

5. Conclusions

Past studies have indicated that wildfire may open the landscape facilitating greater pollen dispersal [20] or potentially decrease pollen transfer through the effects of induced forest edges and changes in pollinator behavior and abundance [18]. Our present study indicates that in northern BC sub-boreal spruce forests, burn severity does not influence pollen abundance in the Asteraceae, Ericaceae, Onagraceae, and Pinaceae families, but does influence the frequency of occurrence of these families across space. Had the wildfire decreased pollen movement, either through changes in the pollinator communities or through dense forest edges acting as barriers to pollen/pollinator movement, we would have expected to see distinct pollen assemblages for severely burned and unburned habitat types. It would have also been expected in this case for the pollen assemblages to be strongly influenced by the immediate plant communities. However, we found no statistically significant differences between burned and unburned pollen abundance. This suggests that the burn matrix of the Shovel Lake wildfire is very heterogeneous, facilitating pollen movement throughout.