A positive relationship between abundance and occupancy of species (AOR) is commonly observed in ecological communities, but the mechanisms driving this pattern are elusive. Succession after disturbance is an important factor structuring many plant communities, yet little is known about how and if this may shape AORs, and if AORs change over time within and between plant communities. Do AORs change through time as a plant community recovers from disturbance? Do patterns in AOR differ between plant communities?
Prince Albert National Park, Saskatchewan, Canada.
Changes in AOR were evaluated through time following prescribed burning of fescue grassland and adjacent forest transition plant communities invaded by Populus tremuloides Michx. using a 35-yr data set. Species presence data were used as a frequency measure of abundance and mean species cover was used as a relative measure for abundance. Species data used for the abundance and occupancy measures were collected in permanent sampling plots prior to burning in 1975, and after burning in 1983, 1995 and 2010.
Changes in grassland AORs based on the frequency were shown through time since disturbance, as an increase in abundance relative to occupancy occurred in 1983 and 2010. There was no significant change in grassland AORs when the relative abundance data were used. The forest transition AOR based on frequency showed the 1983 post-burn AOR had lower abundance relative to occupancy, while the relative abundance measure showed the pre-burn 1975 AOR had lower abundance relative to occupancy. The removal of litter, changes in soil resources and increase in trembling aspen suckering after fire may have caused changes to plant community structure after burning, explaining differing AOR patterns in the grassland and forest transition plant communities.
Although exact mechanisms behind the observed changes in AOR are hard to determine, the phenomenon of change within and between communities over time since disturbance has not previously been documented. The observed differences in AOR between frequency and relative abundance indicate that measures of abundance should be chosen cautiously. Post-disturbance succession should be considered as a mechanism influencing AORs in communities impacted by disturbance.
The tendency for locally abundant species to be widely distributed and locally rare species to be restricted in occurrence is a generally accepted ecological rule (Gaston 1999; Blackburn et al. 2006). The abundance–occupancy relationship (AOR) between the density of individuals of a species within a local area (abundance) and their distribution in space (occupancy) can vary from negative to strongly positive; however, positive relationships have been widely documented over a wide range of taxa and spatial scales (Gaston & Lawton 1990). While previously studied mechanisms may explain some of the observed patterns in AOR, the contribution from each mechanism is variable, depending on the range of species, communities and habitats in a study (Borregaard & Rahbek 2010; Buckley & Freckleton 2010). While many studies have examined how the shape of AOR varies between taxa and communities, few studies have examined how AORs may change over time within communities (Webb et al. 2007). This gap is particularly important, given that in many communities disturbance and subsequent successional patterns are major factors shaping patterns of community composition, biomass, and species abundance and distribution.
Disturbance is a natural component of an ecosystem that maintains ecological resilience through perturbation, directly injuring or killing individuals and indirectly changing resource availability and the physical environment (White & Pickett 1985). Plant community composition and structure is responsive to changes in biotic and abiotic factors after disturbance events (White 1979; Forman 1995). A major role played by disturbance is the deflection of a community from some otherwise predictable successional path (Pickett 1976; Pickett & White 1985). For example, burning may directly remove litter and above-ground biomass, causing changes to soil nutrients, temperature and moisture (Facelli & Pickett 1991). In a forested plant community, increases in soil temperature and loss of apical dominance may promote suckering of woody species (Maini 1960; Ripley & Archibold 1999), negatively affecting low-statured species through shading. However, in a grassland community, burning may promote the growth of low-statured species following the removal of litter (Xiong & Nilsson 1999; Lamb 2008). Although post-disturbance succession is one of the most important processes structuring ecological communities, no studies have examined the role of succession in shaping AORs (Bazzaz 1983; Godron & Forman 1983; Pickett & White 1985; Grime 2001).
While no studies have examined the role of succession in shaping AORs, very few have examined the potential for disturbance and subsequent plant community restructuring to alter AORs. Webb et al. (2007) showed that continual agricultural intensification and habitat modification altered AORs of farmland and woodland birds, with changes to the intraspecific AOR of rare and declining species ultimately driving the form of the community-wide inter-specific AOR. Fisher & Frank (2004) also showed that ongoing anthropogenic disturbance can influence inter-specific AORs in marine fish species. These studies, however, examined gradual modifications from external pressures rather than a disturbance–recovery scenario within a community highly adapted to respond to a disturbance such as fire. Compounding the knowledge gap of the role that post-disturbance succession plays in shaping AORs, few studies have incorporated a sampling period that captures long-term community changes. Stability of AORs through time has been reported in the literature (Buckley & Freckleton 2010), however determining why this stability occurs has not been explored. Successional processes with time since disturbance likely play a role in shaping AORs, however this has been overlooked as a mechanism. This oversight may be because the effects of disturbance, e.g. the removal of litter with burning, are not exclusive of better-studied mechanisms such as resource availability. Disturbance may also have been viewed as a source of noise that could confound or unnecessarily complicate a study. In an effort to control this, sites and systems unaffected by disturbance events during the study period may have been preferentially chosen (Gotelli & Simberloff 1987). As a result, the patterns observed in AOR would be less likely to change due to post-disturbance succession, allowing causation to be attributed to a combination of other mechanisms, such as resource availability, population dynamics and community self-similarity that may act at different spatial scales (Brown 1984; Gaston & Lawton 1990; Hanski et al. 1993; Harte & Ostling 2001; Hubbell 2001; Borregaard & Rahbek 2010). On the other hand, large data sets including sites under varying management and disturbance regimes may have been used without critically assessing if AORs differ among them (Collins & Glenn 1990; Buckley & Freckleton 2010).
Given that succession after disturbance is an important factor structuring plant communities, and that cumulative anthropogenic disturbances can alter AORs (Fisher & Frank 2004; Webb et al. 2007), we expect that AORs may change through time as a community recovers from disturbance. We further expect that grassland and grassland–forest transition plant communities would have different patterns of AOR through time since disturbance. Grassland plant communities are adapted to maintain a stable community structure through the changes in litter and soil resources that follow burning (Xiong & Nilsson 1999; Lamb 2008). Conversely, the resistance and resilience of grassland communities invaded by woody species due to lack of fire disturbance may be degraded and the communities thus more susceptible to changes in community structure following fire (Gunderson 2000; Groffman et al. 2006). As such, we would expect grassland–forest transition AORs to change less over time since disturbance in comparison to grassland AORs.
In this study we examined the role of time since disturbance by fire in shaping the AORs in a remnant Fescue Prairie in Prince Albert National Park, Saskatchewan, Canada. We used a 35-yr data set encompassing a range of times since prescribed burning to assess how AORs changed with post-disturbance succession. We examined two plant communities: an open rough fescue (Festuca hallii (Vasey) Piper) grassland and a grassland–forest transition community dominated by trembling aspen (Populus tremuloides Michx.). The trends observed in AOR differed between plant communities, indicating that while succession after disturbance is important in shaping AORs, differing mechanisms of recovery suggest communities cannot be expected to react in a similar manner. Differences between alternative measures of abundance in the patterns of AOR change over time within communities highlight the importance of both the spatial dynamics of species, and the selection of appropriate spatial scales when assessing AORs. As sampling grid cell size differs with the type of abundance measure used, one must consider the ecosystem, species autecology and plant community response to disturbance in choosing an appropriate abundance measure. In this study, the coarser frequency abundance measure provides clear mechanistic understanding of plant community response in AOR after disturbance that might have been less clear with the relative abundance measure.
The study area is located in the southwest portion of Prince Albert National Park (PANP) (53°36 N, 106°31 W), ca. 50 km northwest of Prince Albert, Saskatchewan, Canada. The study area is within the Boreal Transition Ecoregion, but pockets of the more southern Aspen Parkland Ecoregion are common (Acton et al. 1998). The Aspen Parkland Ecoregion is characterized by forest dominated by trembling aspen (Populus tremuloides Michx.) and other woody species, interspersed with patches of fescue grassland (Cameron 1975; Trottier 1985). One representative fescue grassland in PANP, Wasstrom's Flats, was used for this study. The open grassland in Wasstrom's Flats is dominated by plains rough fescue (Festuca hallii (Vasey) Piper) and awned wheatgrass (Elymus trachycaulus subsp. subsecundus (Link) A. & D. Löve), while forest transition areas surrounding the open grassland and in isolated pockets within it are dominated by trembling aspen. Average 452 mm precipitation occurs at the nearest weather station with long-term data (Big River, SK.; 53°50 N, 107°2 W), with 329 mm as rainfall and 123 mm as snowfall (Environment Canada 2011). Soils are Orthic Black Chernozems occurring on coarse to moderately coarse textured glaciofluvial deposits (Padbury et al. 1978). Current disturbances in the study area include intermittent plains bison (Bison bison L.) and elk (Cervus elaphus L.) grazing and light recreational use from park visitors. Historically fire played a large role in maintaining the grassland area, but fire suppression efforts beginning in ca. 1960 have restricted fire occurrences (Cameron 1975; Gunn et al. 1976; Trottier 1985). Other anthropogenic influences include a vehicle trail and ploughed firebreaks that bisect the site.
Study design and data collection
We examined the association between post-disturbance succession and abundance occupancy relationships using a 35-yr data set (1975–2010) from a long-term prescribed burn study (Gunn et al. 1976; Trottier 1985; Kenkel 2002). The site was initially surveyed in 1975, ca. 28 yrs after the previous known fire. Prescribed burn treatments were conducted over an 8-yr period between 1975 and 1983, with plots subject to either three, four or five burns over the 8-yr period (Table 1). Follow-up surveys of permanent plots were conducted in 1983, 1995 and 2010. The data set thus captures plant community structure prior to burning in 1975, after burning in 1983, and through moderate (1995) and longer (2010) lengths of time for recovery after burning. Sampling was conducted on 16 plots containing a total of 400 permanent quadrats established in 1975. Eight plots were placed in the open grassland community and eight in the adjacent forest transition community. The forest transition community was defined as locations where trembling aspen encroachment was evident in 1975. Each plot contained 25 1 m2 quadrats laid out in a square grid (Fig. 1). Cover of all vascular plant species was recorded in each quadrat during the summers of 1975, 1983, 1995 and 2010. To address potential identification issues stemming from different observers between years, it was necessary to group a small number of species to genus level (Guedo 2012). Individual analyses of prescribed burn treatment effects on AOR were not possible due to the low number of plots exposed to different burn treatments. Two proxy measures of abundance, frequency and relative abundance, were used in this study as determining the number of individuals was not possible with the available data set. The frequency measure of abundance is often applied, given the wide variation in size, growth form and clonality between plant species (Buckley & Freckleton 2010); however, relative abundance measures may be more sensitive to the changes in disturbance observed at the site, and occur on a smaller spatial scale in comparison to frequency (Borregaard & Rahbek 2010). The relative abundance measure may also be advantageous given the uniformity in species composition across the grassland and forest transition plots (Guedo 2012). Using both relative abundance and frequency as measures of abundance allows a unique spatial dynamic to be explored when examining the effects of disturbance on plant community AORs through time. Abundance and occupancy measures for each species in each survey year were derived from species cover data. Frequency measures of abundance were the mean number of quadrats within a plot where a species was recorded (Buckley & Freckleton 2010). Relative abundance measures of abundance were mean species cover over all plots in each respective plant community (Wilson 2008). Occupancy was the number of plots where each species occurred in at least one quadrat. Frequency abundance measures for each species therefore ranged from 0 to 25 given that 25 quadrats were placed in each plot. Relative abundance measures for each species ranged from 0 to 19.72 in the grassland and 0 to 27.76 in the forest transition plant communities. Occupancy of each species ranged from 0 to 8, given that eight plots were placed in the grassland and forest transition plant communities. Occupancy was analysed as the proportion of plots occupied by a species. Abundance and occupancy measures were calculated separately for species occurring in grassland (n = 110 species) and forest transition (n = 118 species) communities. Complete raw data, including species cover data, is archived in the appendices to Guedo (2012).
Table 1. Prescribed burns conducted between 1975 and 1983 during the prescribed burn experiment in Prince Albert National Park. Burn treatment indicates the season and number of burns that took place over the 1975–1983 prescribed burn period, years of burn indicates when the burns took place, and number of grassland and forest transition plots that received the burn treatment
Years of burn
Forest transition plots
5 Autumn burns
1975, 1976, 1979, 1980, 1981
3 Autumn burns
1975, 1979, 1980
4 Spring burns
1976, 1977, 1981, 1982
Differences in inter-specific AORs between years were analysed using generalized linear models using the glm function with a specified binomial error family in the R.2.12.1 package (R Foundation for Statistical Computing, Vienna, Austria). The maximal model used for each plant community–abundance measure combination included occupancy as the continuous response variable, abundance as a continuous explanatory variable, year as a categorical explanatory variable, and the year by abundance interaction term (Table 2). Orthogonal contrasts were used to test a priori predictions of change in AORs over time since disturbance (Crawley 2007). Three orthogonal contrasts allowed either direct or explicit tests for differences in AOR between all years, and were conducted for each plant community–abundance measure; grassland–relative abundance, grassland–frequency, forest transition–relative abundance and forest transition–frequency. Pairings for contrasts were determined by examining model coefficients and examining plots showing separate AORs for each year. Similar AORs between years were then used as starting candidates for contrast analysis. The year by abundance interaction terms in the model were used in orthogonal contrasts, with significant pair-wise comparisons between years indicating significant difference in AOR between those years. Non-significant pair-wise comparisons between years indicated no difference in AOR between those years (Table 3).
Table 2. P-values for change in species occupancy with abundance, year and year by abundance interactions. Columns indicate the respective grassland and forest transition communities, as well as different measures of abundance used: relative abundance and frequency
Table 3. Orthogonal contrasts between year by abundance interactions, indicating differences in abundance occupancy relationships (AOR) over years. As 4 yr of sampling were included in the study, three orthogonal contrasts were possible to compare differences in AOR over years. Columns indicate the respective plant community, the abundance measure used (relative abundance or frequency), the contrasts conducted between years, and z-value and P-values for each contrast
Grassland AORs based on relative abundance did not differ between years, although 1995 was nearly significantly different from other years (Tables 2, 3, Fig. 2a). Grassland AORs based on the frequency measure of abundance differed between years, as contrasts for the year by abundance interaction term showed 1975 and 1995, and 1983 and 2010 were not significantly different, but 1975 and 2010 were significantly different (Tables 2, 3). Thus 1975 and 1995, and 1983 and 2010 followed a similar pattern in AOR (Fig. 2b).
Forest transition AORs based on relative abundance differed between years, as contrasts of the year by abundance interaction terms showed 1975 and 1995, and 1995 and 2010 were not significantly different, but 1975 and 1983 were significantly different (Tables 2, 3). Thus, 1983 followed a separate AOR pattern compared to 1975, 1995 and 2010 (Fig. 2c). Forest transition AORs based on the frequency measure of abundance were also different between years, as contrasts of year by abundance showed 1995 and 2010, and 1995 and 1983 were not significantly different, but 1975 and 1983 were significantly different (Tables 2, 3, Fig. 2d).
The mean abundance (frequency and relative abundance) and occupancy of all species observed across years is shown in Table 4.
Table 4. Mean abundance and occupancy of all species in the grassland and forest transition plots during the prescribed burn study in Prince Albert National Park 1975–2010. The frequency abundance measure (Abun(F)) is the mean number of quadrats within a plot where a species was recorded. The relative abundance measure (Abun(RA)) is mean species cover across all grassland or forest transition plots. Occupancy is the number of plots where each species occurred in at least one quadrat. The mean abundance and occupancy ± 1 SD is provided in each plant community column
2.88 ± 4.57
1.14 ± 2.78
2.83 ± 3.29
2.60 ± 4.14
0.98 ± 2.13
2.68 ± 3.29
5.22 ± 6.63
1.34 ± 2.57
3.79 ± 3.32
4.97 ± 6.46
1.20 ± 1.91
3.96 ± 3.42
4.12 ± 5.74
1.23 ± 2.62
3.58 ± 3.29
3.85 ± 5.30
1.15 ± 2.12
3.58 ± 3.25
6.11 ± 6.94
1.44 ± 1.99
4.48 ± 3.12
4.78 ± 5.99
1.36 ± 3.08
4.12 ± 3.14
As is generally observed in ecological communities, we found positive relationships between abundance and occupancy in the grassland and forest transition communities (Gaston & Lawton 1990; Gaston 1994). The shape of the AOR, however, varied both between and within the plant communities. The AOR changed through time in the grassland community, where the abundance (based on frequency) of species in the 1983 post-burn and 2010 long burn recovery surveys increased relative to occupancy, and was higher than that observed during the 1975 pre-burn and 1995 moderate burn recovery surveys. No significant changes in the AOR were observed in the grassland plant community based on relative abundance measures, indicating sensitivities to the abundance measure used in the grassland community. Changes in AOR were also observed in the forest transition community, where the 1975 pre-burn AOR had lower abundance relative to occupancy when using frequency as a measure for abundance. This contrasts with the AOR based on relative abundance, as the forest transition post-burn 1983 AOR showed a lower abundance relative to occupancy in comparison to other sampling years. These differences in forest transition AOR with abundance measure also suggest sensitivities to the abundance measure used. Differences in the patterns in AOR between grassland and forest transition plant communities demonstrate the susceptibility of AORs to successional change can vary considerably between communities.
Changes to the AOR of a given plant community are likely shaped by the autecology of species, and mechanisms involved with plant community restructuring after disturbance. Grassland community AORs based on frequency changed through time, with the driving mechanism likely interactions between burning, litter cover and soil resources. The removal of litter with burning may cause changes to soil resources and microclimate that can allow space for establishment of an increased number of individuals of some species, or may permit an increase in the growth and cover of plants that were already established. Short-statured graminoids and forbs that may have been suppressed by high amounts of litter such as Carex spp., Koeleria macrantha (Ledeb.) J.A. Schultes, Agrostis scabra Willd., Viola adunca Sm., Cerastium arvense L., Orthocarpus luteus Nutt. and Polygala senega L. showed large increases in abundance relative to occupancy between 1975 and 1983 (Appendix S1). A decrease in abundance relative to occupancy between 1983 and 1995 was observed in low-statured species such as K. macrantha, O. luteus, C. arvense and Carex spp. that had initially increased after burning, and may be due to recovery of litter levels with time since burning. An increase in abundance relative to occupancy between 1995 and 2010 was driven by tall species such as Hierochloe odorata (L.) Beauv. and Potentilla arguta Pursh, species with erect growth forms such as Sisyrinchium montanum Greene, and highly plastic species such as Calamagrostis spp. and Fragaria virginiana Duchesne (Appendix S1). These species may have been able to both grow through high amounts of litter and take advantage soil moisture to increase abundance, as 2010 growing season precipitation was 135% (407 mm) higher than the long-term average (301 mm; Environment Canada 2011). It is also important to note the changes in Solidago species abundance and occupancy across years (Appendices S1–S4). These changes are potentially the result of mis-identification in the 1995 survey. Adhering to criteria that allowed a consistent treatment of the data set, the Solidago spp. were kept separate as the raw data suggested that surveyors in all years were able to identify the species correctly (Guedo 2012). The patterns observed in grassland AORs remained unchanged even when Solidago spp. were grouped, demonstrating these responses are robust to a substantial degree of variation and noise in the data.
The lack of response in grassland AORs based on relative abundance data indicates that at finer scales species cover is relatively consistent, i.e. when present, species with lower or higher cover maintain those trends regardless of disturbance. The nearly significant difference in 1995 AOR may also be due to changes in moisture availability, as the 1995 growing season was 71% (215 mm) of the long-term average (301 mm; Environment Canada 2011). In contrast to the grassland community, AORs in the forest transition community based on relative abundance and frequency both changed in different ways with time since disturbance. While burning likely had similar effects on litter and soil resources in the forest transition community as in the grassland, such effects likely had a much shorter-lived influence relative to the rapid response of the dominant tree species, trembling aspen (P. tremuloides), to burning. Trembling aspen regenerates vigorously from vegetive root suckering following fire, allowing it to regain dominance following disturbance (Maini 1960; Mueggler 1989; Keyser et al. 2005). Burning may create a shorter-lived window for change in species abundance and occupancy in the forest transition compared to the grassland community, regardless of the abundance measure used. The increased abundance relative to occupancy observed in post-burn 1983, 1995 and 2010 with the frequency measure of abundance indicates long-term consistent change in the forest transition plant community after burning. These changes may be due to initial change in environmental conditions through burning, including increased light through the forest canopy, increased soil temperature with reduced litter and forest canopy, and drier soil conditions due to reduced moderating effects of the forest canopy after burning (Fraser et al. 2002). With recovery from burning, however, there may be moderating effects from increased woody species presence and loss of low-statured species. These trends are illustrated by increased abundance (frequency) of low-statured grassland species such as Carex spp., K. macrantha, P. senega and Comandra umbellata ssp. pallida (A. DC.) Piehl between 1975 and 1983, reductions of these species between 1983 and 1995, and large increases in P. tremuloides between 1995 and 2010 (Appendix S3).
The reduced abundance relative to occupancy observed in the post-burn 1983 forest transition community with relative abundance could be explained through the same environmental conditions and species responses. Between pre-burn 1975 and post-burn 1983, abundance (relative abundance) was reduced in larger-growing grasses such as Elymus spp., Bromus spp. and F. hallii, and was accompanied by increases in lower-growing grassland species such as Carex spp., Solidago missouriensis Nutt. and K. macrantha. The abundance of grassland species was reduced and there was an increase in woody species such as Amelanchier alnifolia (Nutt.) Nutt. ex M. Roemer and Rosa acicularis Lindl. between 1983 and 1995, and a large increase in abundance of P. tremuloides in 2010 (Appendix S4).
Positive AORs were observed through time since burning in both grassland and forest transition plant communities. Post-disturbance succession did not alter the overall positive relationship generally seen with AORs, but clearly AORs are dynamic within communities and regularly change over time with recovery from disturbance events. We expect that post-disturbance succession is a contributing mechanism shaping AORs in many plant communities. Further, given that ongoing anthropogenic disturbance is known to change AORs in fish and songbird populations, recovery after disturbance is likely an important mechanism shaping AORs across taxa (Fisher & Frank 2004; Webb et al. 2007). Differences in abundance measures are likely due to differences of scale, where frequency places all species on the same level of assessment as size of individuals is not a contributing factor as it is in relative abundance (species cover). This is especially apparent in grassland ecosystems where there may be large differences in species size, but size may not necessarily reflect dominance within the plant community.
We thank L. Carbyn, G. Trottier, M. Vetter, J. Wilmshurst and those involved in the collection of data, and H. Buckley for assistance with the analysis. Prince Albert National Park of Canada and staff including H. McPhee, G. Rutten, T. Stene, L. Thorpe and J. Weir provided logistical support, and A. Guy assisted in the field. Financial support was provided by an NSERC Discovery grant and a Canadian Foundation for Innovation grant to EGL and a College of Graduate Studies scholarship to DDG.