Research Article |
Corresponding author: Clare E. Aslan ( clare.aslan@nau.edu ) Academic editor: Sven Bacher
© 2020 Clare E. Aslan, Brett G. Dickson.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Aslan CE, Dickson BG (2020) Non-native plants exert strong but under-studied influence on fire dynamics. NeoBiota 61: 47-64. https://doi.org/10.3897/neobiota.61.51141
|
Altered fire regimes are among the most destructive consequences of anthropogenic environmental change. Fires have increased in frequency in some regions, and invasion by fire-adapted non-native species has been identified as a major driver of this change, which results in a feedback cycle promoting further spread by the non-native species and diminishing occurrence of natives. We notice, however, that non-native species are often invoked in passing as a primary cause of changing fire dynamics, but that data supporting this claim are rarely presented. We therefore performed a meta-analysis of published literature to determine whether a significant relationship exists between non-native species presence and increased fire effects and risk, examined via various fire metrics. Our analysis detected a strongly significant difference between fire metrics associated with non-native and native species, with non-native species linked to enhanced fire effects and risk. However, only 30 papers discussing this linkage provided data to support it, and those quantitative studies examined only eight regions, five biome types, and a total of 22 unique non-native taxa. It is clear that we are only beginning to understand the relationship between non-native species and fire and that results drawn from an extremely limited set of contexts have been broadly applied in the literature. It is important for ecologists to continue to investigate drivers of changing fire regimes as factors such as climate change and land use change alter native and non-native fuels alike.
fire extent, fire frequency, fire intensity, flammability, fuels, meta-analysis, native species
Anthropogenic global change has far-reaching consequences. Biodiversity is directly threatened by extinctions (
In places where fire has become more frequent in recent decades, fire regime changes are often the result of non-native species invasions increasing the local density of fine fuels, or of climate change bringing warmer temperatures and increasing the flammability of existing fuels (
Such changes have been shown by multiple metrics to affect fire regimes (
These changes in fire patterns can impact native biodiversity, ecological functions, and ecosystem resilience following disturbances (
Although the link between non-native invasion and problematic shifts in fire is oft-cited in global change literature as an important invasion-fire cycle (e.g.,
To understand how consistently non-natives have been quantitatively associated with increased fire effects and risk, we performed a meta-analysis of published quantitative studies examining the effect of non-native vs native plants on fire characteristics. Our goals were: (a) to determine whether non-native species, relative to functionally similar native species, quantitatively and consistently increase metrics of fire effects and risk in ecosystems, and (b) to gauge the range of contexts over which this has been quantitatively analyzed, in order to consider how broadly assumptions regarding these patterns can justifiably be applied. For this study, we define “fire metrics” as those quantifiable descriptors of fire patterns that can be compared across studies (i.e., fire frequency, fire intensity, flammability, fuels quantity, and fire spatial extent). Note that there have been previous meta-analyses that have examined related but different questions, contributing to our understanding of the link between fire and non-native species.
To perform our meta-analysis, we began by searching ISI Web of Science (with coverage of years 1900–present) to find records of studies that have quantitatively compared non-native and native species’ effects on fire metrics. We used the search terms fire + each of the following: plant + (native* OR exotic* OR non-native* OR alien* OR invasive*); plant + “functional group”; native + (tree* OR shrub* OR perennial grass* OR annual grass*); (severity OR frequency OR intensity OR extent) + (cause* OR attribute*), and applied them to all years inclusive. Additionally, we examined the Literature Cited sections of relevant papers to find additional studies – including from sources not referenced in Web of Science – that might contain relevant quantitative information. Searches were performed in summer 2020.
Although our search terms netted hundreds of references, only papers meeting the following criteria were useful in our meta-analysis: (1) they compared fire metrics stemming from native species (as a control group) with fire metrics stemming from non-native species; (2) they presented comparisons between the metrics of fire associated with native and non-native species from the same plant functional groups; (3) they included quantitative and original fire metrics. Many papers referenced fire effects in discussions of non-native species but did not include original quantitative information. Many other papers examined the effects of fire on non-native species (e.g., reporting experiments examining control measures for non-natives), but we sought the opposite metric: the effect of non-native species on fire. For each of the studies that suited our criteria, we derived from the reports treatment (effect of non-natives) and control (effect of natives) fire metrics as well as sample sizes and measures of variance for treatments and controls.
Across all of the studies we included in our analysis, fire metrics were the response variables of interest. However, there are many ways to measure the effect of a given factor (e.g., non-native fuels) on fire. We were able to include all of these in one common meta-analysis framework by using the ratio of means (ROM) to compare the treatment and control effects of all studies (
Categorical analysis can be used to further explore the population of studies included in the overall meta-analysis in order to ascertain whether significant treatment effects persist within certain limited contexts. As long as a given category is represented by at least two studies, it is possible to examine it separately from the other categories to measure the strength of the treatment effect within that categorical context. The categories we examined as such included: biome type, geographic region, plant functional group, and fire effect metric. Because the total number of studies within each category was not always greater than 1, the total number of studies included in categorical analyses did not always equal the total number of studies included in the overall meta-analysis.
In meta-analyses, studies included in the calculation of effect sizes are weighted more heavily if they used a larger sample size in the original research. We included the variances and sample sizes of all studies in our response ratio meta-analysis by calculating fixed and random effects estimates and applying inverse variance weighting, thus allowing studies with larger sample sizes to carry greater influence on the effect estimates. We calculated heterogeneity Q statistics to evaluate whether effect sizes are homogeneous or, conversely, are suggestive of underlying unexplained structure in the data (
An important consideration in meta-analysis is that researchers and journals may be less inclined to publish studies that fail to show the expected effect, either because the results were non-significant (in our case, finding no difference between native and non-native species and their effects on fire metrics) or because they were significant in the opposite direction from predicted (in our case, finding that native species enhanced fire metrics more than non-native species). To estimate the potential quantitative effect of this phenomenon, we calculated a fail-safe analysis, which we performed using the trimfill function in the meta R package. Results indicated whether the outcome of our overall meta-analysis was likely affected by a lack of publication or “file drawer” problem and also estimated the likely overall effect size after producing a correction for such a publication bias. Note that sample sizes were not sufficient to conduct a similar fail-safe analysis for subgroup categories.
Our search terms yielded 612 unique sources. We examined each of these for methodology and found only 30 papers, reporting results of 41 distinct studies, that included a usable quantitative comparison of the effects of native vs non-native species on fire metrics. This final sample of relevant papers displayed the following breakdown by subgroup categories: by region, nine studies took place in the Southwestern US, three in mediterranean California, five in the Western US more broadly, eight in Australia, six in the Eastern US, one in Europe, three in South Africa, and six in South America. By biome, six studies were performed in deciduous forest, 10 in desert, 16 in mediterranean systems, eight in savanna, and one in mixed shrubland/woodland. Functional groups included annual grasses (9 studies), forbs (1 study), perennial grasses (20 studies), shrubs (3 studies), and trees (6 studies). A total of five usable studies combined data from multiple species to report fire metrics of non-native vs native species, making it impossible to extract the contributions of individual species but allowing comparison between those two groups.
Overall, our meta-analysis detected a strong, statistically significant link between non-native species and increased fire effect (random-effects model ROM = 2.21; 95% bias-corrected CI 1.52 to 3.20; n = 41; p < 0.0001). Heterogeneity was also significant (QT = 5.68 × 105, df = 40, p < 0.0001), which highlights the large amount of unexplained data structure in the dataset. The many different approaches to comparing fire stemming from natives and non-natives that were employed by the various studies we examined likely contributed to this heterogeneity, emphasizing the importance of subgroup comparisons. The relevant studies found for this analysis contained clear evidence of a link between non-native species and enhanced fire metrics. However, the total number of species and the total number of contexts covered is extremely limited. The 41 studies reported species-level fire metrics reported for only 16 taxa (Table
The 16 species examined at the species level in quantitative comparisons of fire metrics stemming from native vs. non-native fuels, and key traits related to effects. Five analyzed studies compared groups of native vs non-native species and thus effects could not be ascribed to individual species, and these studies are excluded from this table.
Species | Traits related to fire effects in meta-analysis | Citation |
---|---|---|
Ampelodesmos mauritanica | Resprouts quickly after fire; produces flammable biomass more rapidly than native species |
|
Andropogon gayanus | High growth potential relative to native species |
|
Bromus hordeaceus | Low quality litter decomposed less than native litter, contributing to regional fuels for a longer period of time; compared with native species, sustains dry biomass for a larger portion of the year |
|
Bromus rubens | Winter annuals that escape extreme summer heat, generating high fuel load production relative to native species |
|
Bromus tectorum | Exploits soil water following fire, outcompeting natives in regeneration |
|
Cenchrus ciliaris | Increases fuel loads relative to native species |
|
Cytisus scoparius | Higher relative growth rate than native species |
|
Eragrostis lehmanniana | Much faster biomass production than native species |
|
Hakea sericea | Increased fuel loads relative to native species |
|
Hyparrhenia rufa | Higher growth rates in fertile sites, relative to native species |
|
Imperata cylindrica | Increased fuel loads and fuel continuity relative to native species |
|
Melinus minutiflora | Higher growth rates in fertile sites, relative to native species |
|
Pennisetum setaceum | Increased fuel loads relative to native species |
|
Pinus contorta | Increased vertical fire continuity and increased flammability of fuels relative to comparison native species |
|
Schinus terebinthifolius | Reduces fire frequency ecosystem-wide |
|
Tamarix sp. | Rapid biomass accumulation and rapid regrowth after fire |
|
Region | Biome type | Study |
---|---|---|
Southwestern US | Savanna |
|
Desert |
|
|
|
||
|
||
|
||
|
||
|
||
Western US | Savanna |
|
|
||
|
||
Desert | Melgoza et al. 1989 | |
|
||
Australia | Mediterranean |
|
|
||
|
||
|
||
|
||
Eastern US | Deciduous forest |
|
|
||
|
||
Europe | Mediterranean |
|
California | Mediterranean | Keeley 2001 |
|
||
South Africa | Mediterranean |
|
|
||
South America | Savanna |
|
|
The significant response ratio of non-natives to natives held across almost all examined subgroups, as well. Among metrics of fire effects and risk, non-natives generated significantly higher fire metrics of flammability (random-effects model ROM = 1.50; 95% bias-corrected CI 1.39 to 1.62; n = 15; QB = 1923.69), fuels (ROM = 2.27; CI 1.30 to 3.97; n = 18; QB = 6.48 × 104), and spatial extent (ROM = 10.02; CI 3.19 to 31.48; n = 2; QB = 30.30) (Fig.
Ratios of means (dark circles) and 95% confidence intervals (denoted by lines) for fire metric subgroups analyzed using Hedges’ d response ratios. Positive means and confidence intervals excluding 1 (indicated by a dashed, horizontal line) can be considered to indicate significantly higher fire metrics for native than non-native species. Sample sizes of each subgroup are denoted with numerals above each line.
Ratios of means (dark circles) and 95% confidence intervals (denoted by lines) for biome subgroups analyzed using Hedges’ d response ratios. Positive means and confidence intervals excluding 1 (indicated by a dashed, horizontal line) can be considered to indicate significantly higher fire metrics for native than non-native species. Sample sizes of each subgroup are denoted with numerals above each line.
Among functional groups, non-natives generated significantly higher fire metrics for perennial grasses (ROM = 2.53; CI 1.55 to 4.10; n = 20; QB = 4.66 × 104), shrubs (ROM = 1.41; CI 1.34 to 1.49; n = 3; QB = 0.27), trees (ROM = 1.73; CI 1.51 to 1.99; n = 6; QB = 1695.02), and annual grasses (ROM = 2.39; CI 1.24 to 4.60; n = 9; QB = 1. × 104) (Fig.
Ratios of means (dark circles) and 95% confidence intervals (denoted by lines) for functional group subgroups analyzed using Hedges’ d response ratios. Positive means and confidence intervals excluding 1 (indicated by a dashed, horizontal line) can be considered to indicate significantly higher fire metrics for native than non-native species. Sample sizes of each subgroup are denoted with numerals above each line.
Ratios of means (dark circles) and 95% confidence intervals (denoted by lines) for region subgroups analyzed using Hedges’ d response ratios. Positive means and confidence intervals excluding 1 (indicated by a dashed, horizontal line) can be considered to indicate significantly higher fire metrics for native than non-native species. Sample sizes of each subgroup are denoted with numerals above each line.
The fail-safe analysis detected a significant file drawer problem (t = −2.49; df = 39; p = 0.017). Upon correcting for this problem via an estimate of the effect size in the absence of a file-drawer problem, the trimfill recommended analysis assumed an addition of 19 non-significant studies but predicted that with such studies included the overall effect would remain significant (ROM = 5.46; CI 3.83 to 7.77; p < 0.0001) and that the link between non-native species and increased fire metrics would persist.
Our meta-analysis found a significant link between non-native plant species and fire metrics broadly, and specifically found that non-natives are associated with increased fuels, fire intensity, flammability, and fire extent, compared with native plant species, where the two have been contrasted. That is, quantitative research finds evidence that non-natives alter fire regimes by shifting the characteristics, quantity, and/or flammability of fuels. At the same time, our search terms netted a very small number of quantitative studies examining an even smaller number of species. Approximately a twentieth of the studies that met our search criteria compared native and non-native plant species quantitatively; most of the others simply referenced the relationship between non-natives and fire.
Together, these results suggest that we are only beginning to understand the role of non-native species in fire regimes under environmental change, globally. Effects are quite strong where they have been quantitatively analyzed, but analyses have been limited to a few contexts. By one estimate, there are almost 17,000 species that have been established outside their native range (
The 16 taxa that were examined at a species level in these studies included annual and perennial grasses, trees, shrubs, and a forb. Nearly all of them contributed to increases in fire metrics; the sole exception was Schinus terebinthifolius, which is associated with decreased fire frequency that promotes further invasion by this non-native tree (
Non-native plants have led to novel fire disturbances in the studied systems. Our results demonstrated that fires in invaded sites recur with greater frequency, burn with higher fireline intensity, or burn over greater extent than in native-dominated sites. Such systems are subject to significant ecological transformation: when native species are non-fire-adapted or unable to recover from severe fires, a positive invasive species-fire feedback cycle emerges (
Altered fire regimes are evidently a strong component of global change. Furthermore, they interact with other drivers of environmental change. Climate change alone can boost the growth rate (i.e., production and thus total quantity of fuels) and flammability of biomass both native and non-native (
Disentangling the effects of climate change, land use, and non-native fuels will be important for spatially-explicit fire risk assessment and management and restoration decision-making (
Global change today consists of multiple drivers operating both individually and in synergy. The combined influence of biological invasions, land use change, and climate change can result in dramatic changes in fire dynamics within particular systems, yet understanding how each driver contributes to fire regime change is essential for effective management decision-making in response. Our study identified a clear role of non-native species in increased fire metrics, but also highlighted the limited scope of our understanding – only a small number of species and systems have been quantitatively examined to this point. Both native and non-native fuels must be considered in light of changing climatic patterns and land uses, and increased empirical assessment of the respective roles of climate, land use, and invasion are necessary for appropriate responses.
We thank M. Gray, L. Zachmann, C. Levine, and two anonymous reviewers for key discussions and thoughtful reviews of early drafts of this manuscript. This work was supported by Joint Fire Science Program Award #15.232.