Increasing and declining native species in urban remnant grasslands respond differently to nitrogen addition and disturbance

Abstract

Background and Aims

Atmospheric nitrogen deposition and natural fire regime suppression are key drivers of vegetation change in urbanizing grasslands. Some species thrive under these conditions, while others face local extinction. In the natural grasslands that surround Melbourne, Australia, biotic homogenization has occurred with intensifying urbanization. Some native species have become rarer (decreaser species) across the landscape, while others have become more widespread (increaser species). This study experimentally examined the response of increaser and decreaser plant species to nitrogen addition/depletion, and examined the presence/absence of annual disturbance to the vegetation.

Methods

Decreaser and increaser species were planted into 60 field plots established in an urban Melbourne grassland and examined over 2 years. Annual removal of above-ground biomass occurred in half the plots to simulate biomass removal via fire, with the remaining plots undisturbed. Soil nitrogen was depleted in one-third of plots, one-third received no nitrogen treatment and one-third were fertilized with nitrogen. Increaser plant species were predicted to persist in the absence of disturbance, and thrive when fertilized. In contrast, high mortality was predicted for decreaser species in the absence of disturbance, with fertilization providing no advantage.

Key Results

Seedling mortality for increaser and decreaser species was unrelated to the treatments. The mortality of decreaser species was high (69 %), and the mortality of increaser species low (20 %). However, seedling growth was related to the treatments. The total biomass of decreaser species was highest in annually disturbed plots, with growth suppressed in undisturbed plots. In contrast, the total biomass of increaser species was unrelated to the disturbance regime, but responded positively to nitrogen enrichment.

Conclusions

The results provide evidence that by affecting plant growth, declines in biomass removal and atmospheric nitrogen deposition could be key drivers of biotic homogenization in urban grasslands.

INTRODUCTION

Urbanization is a major driver of biodiversity change (McDonald et al., 2008; Secretariat of the Convention on Biological Diversity, 2012). Remnant vegetation that becomes fragmented in urban landscapes is exposed to novel environmental conditions (Pickett et al., 2001; Grimm et al., 2008, Kowarik, 2011), with these conditions driving declines in the abundance and geographical range of some species, and increases in others (Vellend et al., 2017). McKinney and Lockwood (1999) termed these species ‘winners’ and ‘losers’, with the number of losers greatly outnumbering the number of winners. This disparity in the number of winners and losers is considered a major cause of biotic homogenization (McKinney and Lockwood, 1999), as the globe’s ecosystems become more similar to each other at community (Clavel et al., 2010) and regional scales (Kuhn and Kultz, 2006; McKinney, 2006).

Remnant vegetation in urban landscapes is typically exposed to elevated nitrogen levels because of atmospheric deposition, which is exacerbated relative to background levels because of emissions due to industry and transport (Zhu et al., 2006; Fan et al., 2014). Elevated soil nitrogen increases plant productivity (Stevens et al., 2015), and nitrogen can accumulate rapidly in former agricultural soils at the urban fringe (Raciti et al., 2011). However, elevated nitrogen can also have negative effects for some species (Zhang et al., 2016; Caplan et al., 2017). In natural grasslands, elevated nitrogen can increase the ratio of non-native to native plant diversity (Stevens et al., 2015), and shift species composition to plants with a novel representation of functional traits (La Pierre and Smith, 2015).

For urbanized native grasslands in south-eastern Australia, periodic disruption to productivity through natural disturbance regimes has been reduced via fire suppression (Williams et al., 2005a, b). Like small isolated vegetation communities elsewhere (e.g. Bond et al., 2005; Ramalho et al., 2014; Pickens et al., 2017), a reduction in historical fire frequency has resulted in a decline in vegetation condition (Williams et al., 2005a). Grassland diversity on the basalt plains of south-eastern Australia is vulnerable to biomass accumulation over short time periods (Morgan, 2001). In the absence of disturbance, dominant tussock grasses develop a dense canopy cover (Morgan and Lunt, 1999), reducing light transmission into inter-tussock spaces (Morgan, 1998b), starving the inter-tussock flora of light and recruitment space (Morgan, 1998b, c). Thus, with increasing time since disturbance, the abundance and richness of inter-tussock flora declines (Lunt, 1994; Morgan, 1999). In combination with elevated soil nitrogen, the effect may be synergistic, with biomass accumulation being both undisrupted and accelerated.

In the natural grasslands that surround the city of Melbourne, Australia, biotic homogenization has been occurring (Zeeman et al., 2017). The past two decades has been a period of intense urbanization in Melbourne, and has seen the proportion of species shared between remnant grassland sites increased (Zeeman et al., 2017). Over this period there have been increases in the geographical range of some species (winners) and declines for others (losers; Zeeman et al., 2017).

To further our understanding of how urbanization affects floristic change in Melbourne’s grasslands, we examined a group of native plant species that have become more widespread over this two-decade period, and a group that have declined in range. The availability of soil nitrogen, and lack of biomass removal through disturbance, was experimentally altered in the field to examine whether greater access to nitrogen and the absence of disturbance favours ‘increasing species’, and disadvantages ‘declining species’. We predicted that increaser species would persist in the absence of disturbance and thrive under enhanced nitrogen conditions. In contrast, we predicted that the mortality of decreaser species would be high in the absence of disturbance and that greater access to nitrogen would provide no advantage.

MATERIALS AND METHODS

Study site

This study was undertaken at the Iramoo Wildflower Grassland Reserve in Melbourne, Australia (Supplementary Data S1). The site receives a mean annual rainfall of 540 mm, with monthly mean maximum temperature peaking in January at 25.7 °C, and dropping to 13.7 °C in July. The 40-ha reserve is dominated by the indigenous C₄ grass Themeda triandra but, like many of Melbourne’s native grasslands, the reserve also contains areas dominated by invasive exotic grasses. The experiment was established in a section of long-unburnt (>15 years) Themeda-dominated grassland (Supplementary Data S1).

Experimental design

The experiment ran from August 2014 to October 2016. At the commencement of the experiment, 60 1-m² grassland plots were randomly selected within the study site. Thirty plots were then randomly assigned as the ‘annual disturbance plots’ and were brush cut to ground level; all cut biomass was removed. Brush cutting of the ‘annual disturbance’ plots was repeated in April 2015 and April 2016. The remaining 30 plots remained uncut for the duration of the experiment, designated as the ‘no disturbance’ plots.

Using data presented by Zeeman et al. (2017) (Supplementary Data S2), eight native herbaceous/sub-shrub species that had increased in site-level frequency in Melbourne’s grasslands over the past two decades, and eight native species that had declined in frequency, were selected for the study (Table 1). These species differed in the composition of functional traits (Table 1). For each species, a total of 30 tubestock seedlings were used (tubes measured 50 mm in width × 120 mm in depth). Within each of the 60 plots, four increaser and four decreaser species were randomly allocated and planted amongst the existing grass tussocks, providing a total of eight seedlings in each plot. Roofing nails were placed next to the seedlings to enable detection over time.

Table 1.

Functional trait data sourced from a database held in the Department of Ecology, Environment and Evolution, La Trobe University

Species	Family	Raunkiar life-form	Height (cm)	Seed mass (mg)	SLA (mm2 mg⁻1)
Increaser species
Atriplex semibaccata	Amaranthaceae	Chamaephyte	50	0.7	120
Calocephalus lacteus	Asteraceae	Chamaephyte	70	0.58	12
Convolvulus angustissimus	Convolvulaceae	Protohemicryptophyte	10	8.21	21
Einadia nutans	Chenopodiaceae	Chamaephyte	50	0.78	55
Enchylaena tomentosa	Amaranthaceae	Chamaephyte	100	12.18	66
Senecio hispidulus	Asteraceae	Hemicryptophyte (partial)	80	0.05	20
Senecio quadridentatus	Asteraceae	Hemicryptophyte (partial)	110	0.19	25
Veronica gracilis	Scrophulariaceae	Chamaephyte	60	0.06	25
Decreaser species
Dichondra repens	Convolvulaceae	Protohemicryptophyte	5	1.61	23
Eryngium ovinum	Apiaceae	Protohemicryptophyte	60	1.93	7
Glycine tabacina	Fabaceae	Protohemicryptophyte	20	8.08	6
Goodenia pinnatifida	Goodeniaceae	Hemicryptophyte (partial)	25	3.27	7
Hypericum gramineum	Hypericaceae	Protohemicryptophyte	25	0.01	22
Leptorhynchos squamatus	Asteraceae	Protohemicryptophyte	20	0.08	20
Lobelia pratioides	Campanulaceae	Protohemicryptophyte	7	0.10	21
Solenogyne dominii	Asteraceae	Hemicryptophyte (flat)	5	0.32	22

The seedlings of these species were planted in a Melbourne grassland, with nitrogen availability and disturbance frequency experimentally altered. SLA, specific leaf area.

Within the 30 ‘annual disturbance’ and 30 ‘no disturbance’ plots, a further three sub-treatment groups were established. Ten plots in each group experienced nitrogen enrichment, with urea added to the plots every 4 months at a rate of 5 g per 1-m² plot throughout the duration of the experiment, equating to 150 kg ha⁻¹ year⁻¹. This is likely to be higher than ambient rates, but as others have done (e.g. Borer et al., 2013; Morgan et al., 2016), applying higher loads is an effective method of forcing a short-term response within a study period. A second sub-group of ten plots in each group experienced nitrogen depletion; following Prober and Lunt (2009), 0.5 kg of sucrose (white sugar) was sieved over each 1-m² plot every 3 months throughout the duration of the experiment. The third sub-group acted as a control, receiving no nitrogen or sucrose treatment. Plots received 9 L of water after planting, with watering then continuing at a rate of 4.5 L per week for 6 months after planting (August to January) to ensure initial seedling survival.

In total, our study created six treatment regimes: (i) annual disturbance/N depleted, (ii) annual disturbance/no N treatment, (iii) annual disturbance/N-enriched, (iv) no disturbance/N depleted, (v) no disturbance/no N treatment and (vi) no disturbance/N-enriched (Supplementary Data S3). Seedling mortality was monitored annually in each of the plots in October 2014, 2015 and 2016. Following the final survey in October 2016, all surviving seedlings were carefully dug out and removed with tap root intact, and the biomass from all plots was cut, collected and sorted according to native and exotic status. All removed seedlings and biomass was oven-dried at 70 °C for 48 h and then weighed. Soil pH was recorded in every plot, with the expectation that N-enriched plots would have lower pH than N-depleted plots given the acidification effect of N fertilization (Barak et al., 1997; Guo et al., 2010).

Data analysis

Generalized linear models were used to assess the mortality of increaser and decreaser species over the duration of the experiment. A binomial distribution was used with complementary log-log as the link function. Differences in soil pH and plant biomass in the final year (total biomass, native biomass, non-native biomass, combined above- and below-ground seedling biomass) was assessed using one-way ANOVA. Where significant differences were identified between treatment groups, an least significant difference (LSD) post-hoc analysis was undertaken to identify which groups were driving the differences. Analyses were undertaken in the statistical package SPSS (Version 21, IBM Corp., Armonk, NY, USA) with significance identified at an α = level of 0.05.

RESULTS

Soil pH and plot biomass

Both the annual disturbance/N-enriched and the undisturbed/N-enriched plots had significantly lower soil pH than all other plots (F_5,54 = 12.41, P < 0.001, Fig. 1A) (see Supplementary Data S4 for post-hoc analysis). Significant positive treatment effects of nitrogen addition and the absence of disturbance were identified on total plant biomass (F_5,54 = 12.98, P < 0.001, Fig. 1B), native biomass (F_5,54 = 12.03, P < 0.001, Fig. 1C) and non-native biomass (F_5,54 = 12.41, P < 0.001, Fig. 1D).

Fig. 1.

The effect of nitrogen and biomass removal treatments on (A) soil pH, (B) total plant biomass, (C) native biomass and (D) non-native biomass, at the completion of the 3-year experiment in an urban Melbourne grassland. Experimental plots were divided evenly into six treatment groups: AD/ND = annual disturbance/N-depleted; AD/NNT = annual disturbance/no N treatment; AD/NE = annual disturbance/N-enriched; ND/ND = no disturbance/N-depleted; ND/NNT = no disturbance/no N treatment; ND/NE = no disturbance/N-enriched. Error bars represent 1 s.e. Different letters between groups represents statistically significant differences. Groups with the same letters are not significantly different.

Both total plant biomass and non-native biomass were significantly lower for annual disturbance/N-depleted plots than all treatments other than the annual disturbance/no N treatment plots (see Supplementary Data S5 and S6 for post-hoc analysis). Differences between the annual disturbance/no N treatment, annual disturbance/N-enriched and undisturbed/N-depleted plots were non-significant. However, the undisturbed/no N treatment and undisturbed/N-enriched plots had significantly higher levels of biomass than all other plots (see Supplementary Data S5 and S6 for post-hoc analysis). The undisturbed/N-enriched plots also had significantly higher levels of native biomass than both the annually disturbed/N-depleted and the annually disturbed/no N treatment plots (see Supplementary Data S7 for post-hoc analysis).

Seedling mortality

Both decreaser and increaser species experienced significant mortality during the experiment (declining species: β = −1.48, s.e. = 0.14. P < 0.001; increasing species: β = −0.99, s.e. = 0.17, P < 0.001). However, mortality was unrelated to treatment (declining species: P = 0.37; increasing species: F = 0.75, P = 0.57; Fig. 2A, B, see Supplementary Data S8 for β coefficients).

Fig. 2.

The effect of nitrogen enrichment and the absence of biomass removal on the mean (±s.e.) survival of native plant species grown in urban grassland plots in Melbourne, Australia, over a 3-year period. (A) ‘Decreaser species’, (B) ‘increaser species’ and (C) overall mortality across treatments. Increaser species included: Atriplex semibaccata, Calocephalus lacteus, Convolvulus angustissimus, Einadia nutans, Enchylaena tomentosa, Senecio hispidulus, Senecio quadridentatus and Veronica gracilis. Decreaser species included: Dichondra repens, Eryngium ovinum, Glycine tabacina, Goodenia pinnatifida, Hypericum gramineum, Leptorhynchos squamatus, Lobelia pratioides and Solenogyne dominii.

Overall, 78 ± 3 % of decreaser plants survived the first year, bu, only 31 ± 4 % survived the second year. In contrast, the mortality of increaser species was significantly lower (F = 89.55, P < 0.001), with 98 ± 1 % surviving the first year, and 80 ± 3 % surviving the second year (Fig. 2C).

Seedling growth

After 2 years, the combined above- and below-ground biomass of planted seedlings in each plot differed significantly between treatments for both the decreaser (F_5,54 = 4.53, P = 0.002) and the increaser species (F_5,54 = 6.32, P < 0.001). The combined above- and below-ground biomass of decreaser species was significantly higher in annually disturbed plots than in undisturbed plots (Fig. 3) (see Supplementary Data S9 for post-hoc analysis). In contrast, the increaser species had significantly higher combined above- and below-ground biomass in the undisturbed/N-enriched plots compared to all other plots (Fig. 3) (see Supplementary Data S10 for post-hoc analysis).

Fig. 3.

The effect of nitrogen and biomass removal treatments on mean (±s.e.) surviving biomass of native ‘increaser species’ and native ‘decreaser species’ per plot after 3 years of growth in an urban grassland in Melbourne, Australia. Plots were divided evenly into six treatment groups: AD/ND = annual disturbance/N-depleted; AD/NNT = annual disturbance/no N treatment; AD/NE = annual disturbance/N-enriched; ND/ND = no disturbance/N-depleted; ND/NNT = no disturbance/no N treatment; ND/NE = no disturbance/N-enriched. Increaser species included: Atriplex semibaccata, Calocephalus lacteus, Convolvulus angustissimus, Einadia nutans, Enchylaena tomentosa, Senecio hispidulus, Senecio quadridentatus and Veronica gracilis. Decreaser species included: Dichondra repens, Eryngium ovinum, Glycine tabacina, Goodenia pinnatifida, Hypericum gramineum, Leptorhynchos squamatus, Lobelia pratioides and Solenogyne dominii.

DISCUSSION

We investigated how declines in fire frequency and soil nitrogen enrichment drive floristic change in urbanized natural grasslands. Such mechanistic insight is necessary to understand the observed trajectories of grassland plants in urban grasslands (Zeeman et al., 2017). We planted seedlings of native species that had become more common in Melbourne’s grasslands over the past two decades, and seedlings of species that had become less common. We then manipulated disturbance frequency and the availability of soil nitrogen over 2 years, seeking to examine whether these factors could be driving observed floristic changes.

Mortality for species which have become more common in the landscape was low irrespective of disturbance frequency or nitrogen availability, while mortality was high for the species which have become less common. These two groups of species differed in the composition of functional traits, with increaser species on average have taller growth form, higher specific leaf area and higher seed mass (Supplementary Data S11).

Treatments significantly influenced the growth of seedlings, but in different ways. For the increaser species, enhanced access to soil nitrogen greatly promoted growth in the absence of disturbance. For decreaser species, growth was unaffected by nitrogen availability, with growth highest when biomass was annually removed. This result suggests the two species groups differ in basic resource needs. While the increaser species are limited by soil resources, decreaser species are limited by light availability.

Disturbance effects

Melbourne is currently undergoing a period of rapid urbanization (Gleeson and Spiller, 2012; OECD, 2012). This has coincided with declines in fire frequency in the critically endangered native grasslands now being transformed by urbanization with resultant biotic homogenization of the flora (Williams et al., 2005a, b; Zeeman et al., 2017). Declines in the historical frequency of fire has been identified as a key driver of floristic change for grasslands in the United States (e.g. Briggs and Knapp, 2001), South Africa (e.g. Govender et al., 2006) and Australia (e.g. Lunt et al., 2012). In the absence of fire, grass biomass accumulates, restricting the ability for many herbaceous plants to access light and space (Morgan and Lunt, 1999). Ultimately, this leads to their competitive exclusion (Lunt, 1994; Morgan, 1999), and in the case of grasslands in south-eastern Australia, irreversible changes due to the transient nature of soil seedbanks (Morgan, 1998a). However, our data suggest that not all native species are vulnerable to a lack of biomass removal, with some native species becoming more common in Melbourne’s grasslands with declining fire frequency (Zeeman et al., 2017). Thus, a lack of biomass removal was a key line of investigation when examining the mechanism of biotic homogenization in Melbourne’s grasslands.

Although a relationship between seedling mortality and disturbance was not identified, we did find a relationship between seedling growth and disturbance, which may affect fitness at later life stages such as flowering and seed production (Lunt, 1994). The total surviving biomass of decreaser species was significantly higher in plots that experienced annual disturbance, indicating that in the absence of biomass removal, these species experienced growth suppression. Furthermore, decreaser species made a greater contribution to total biomass than increaser species in disturbed plots despite being on average smaller plants (Zeeman et al., 2017), and having fewer surviving individuals. Thus, the surviving seedlings of decreaser species appeared to possess a more vigorous re-sprouting ability relative to the increaser species.

Nitrogen effects

The concentrated burning of fossil fuels that occurs in urban environments is a cause of atmospheric nitrogen deposition (Brazel et al., 2000; Pickett et al., 2001; Grimm et al., 2008). Given that Melbourne is a rapidly expanding city (Gleeson and Spiller, 2012; OECD, 2012), the potential effect of nitrogen fertilization on increaser and decreaser plant species informed our second line of investigation.

As expected, nitrogen addition significantly enhanced plot-level biomass production, while nitrogen depletion (via addition of sucrose) suppressed production. However, we found little effect of nitrogen manipulation on the survival or growth of the decreaser species. Like many Australia plants that do not respond to fertilizer addition (Morgan et al., 2016), these species appear to operate at low resource levels. In contrast, the increaser species grew significantly larger in the fertilized plots when disturbance was absent. This result extends experiments examining the effect of nitrogen addition on grasslands elsewhere. For example, in a global grassland experiment, Stevens et al. (2015) identified that nitrogen addition favours non-native plants over native plants. Here, we demonstrate that there is a portion of the native flora that also demonstrate a positive response, and that these same species are those that are increasing in the urban landscape.

McDonnell and Hahs (2015) described native species that thrive under urban conditions as pre-adapted, able to thrive through a unique combination of phenotypic traits and/or plasticity. The increaser species in Melbourne’s grasslands appear to be pre-adapted to the urban landscape, thriving under enhanced nitrogen conditions and unaffected by the accumulation of biomass. By contrast, the decreaser species appear maladapted, unable to gain benefit from nitrogen addition and significantly supressed under the accumulation of biomass. Hence, we have (the first) evidence for native plants responding differently to urbanization, and the underlying reasons for this response.

Recruitment dynamics

In our study, recruitment dynamics were not examined as all plants used in the experiment came from tube-stock. However, recruitment is a key process in need of further investigation. Presumably, decreaser species face significant barriers to recruitment in the absence of disturbance. Under dense grass cover, flowering is likely to be suppressed, and opportunities for seedling establishment restricted. However, due to the tall growth form of increaser species (Zeeman et al., 2017), light is accessible to these species, and, with buds developing above the grass canopy, flowering, pollination and dispersal can occur unrestricted. Further research is needed at the germination and early growth stages when both the increaser and the decreaser species will need to endure poor light conditions. However, there is evidence from Melbourne’s grasslands that increaser non-native Asteraceae species are capable of germinating under dense grass cover (Lunt and Morgan, 1999); thus, the increaser native flora may demonstrate similar capabilities to the increaser non-native flora.

CONCLUSIONS

Overall, our experiment was unable to demonstrate that nitrogen enrichment and the absence of disturbance cause higher transplant mortality for the species that are declining in Melbourne’s grasslands, compared to those that are increasing. This may be because it is the germination phase that is most vulnerable. However, we did demonstrate a differential treatment effect on the growth of increaser and decreaser species. The species that are declining in Melbourne’s grasslands can re-sprout vigorously following disturbance relative to the increaser species. In the absence of disturbance, however, the decreaser species experience significant growth suppression, which is likely to impact on their ability to recruit.

The species that are increasing in Melbourne’s grasslands are robust, demonstrating low mortality irrespective of disturbance frequency and nitrogen availability. In addition, under nitrogen-enriched conditions, the increaser species are capable of vigorous growth in the absence of disturbance. Our results provide evidence that by effecting plant growth, declines in historical fire frequency and atmospheric nitrogen deposition could be key drivers of biotic homogenization in Melbourne’s grasslands. As Melbourne becomes increasingly urbanized, the remnant grasslands that exist within its urban boundary are likely to undergo continuing change, with a suite of native species pre-adapted to the urban environment thriving.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Supplementary Data S1: Study site at the Iramoo Wildflower Grassland Reserve in Melbourne, Australia. Map illustrates the design of the experiment. Supplementary Data S2: The site-level occurrence frequency of native herbaceous and sub-shrub species in Melbourne’s grasslands. Supplementary Data S3: Experimental plots in October 2016. Supplementary Data S4: Significance of differences in soil pH between treatments. Supplementary Data S5: Significance of differences in total biomass between treatments. Supplementary Data S6: Significance of differences in exotic biomass between treatments. Supplementary Data S7: Significance of differences in native biomass between treatments. Supplementary Data S8: Generalized linear regression beta regression coefficients and significance values for plant mortality across treatments. Supplementary Data S9: Significance of differences in declining species biomass between treatments. Supplementary Data S10: Significance of differences in increasing species biomass between treatments. Supplementary Data S11: Boxplots demonstrating differences in the spread of trait values for (a) plant height, (b) seed mass and (c) specific leaf area for increaser and decreaser species used in the experiment.