Biogeochemistry

, Volume 109, Issue 1, pp 63–83

Soil microbial responses to fire and interacting global change factors in a California annual grassland

Authors

  • Kathryn M. Docherty
    • Department of Biological SciencesWestern Michigan University
    • Center for Ecology and Evolutionary Biology (CEEB)University of Oregon
  • Teri C. Balser
    • Department of Soil ScienceUniversity of Wisconsin-Madison
    • Department of Soil and Water ScienceUniversity of Florida
  • Brendan J. M. Bohannan
    • Center for Ecology and Evolutionary Biology (CEEB)University of Oregon
    • Department of Soil ScienceUniversity of Wisconsin-Madison
    • Department of Soil EcologyHelmoltz-Centre for Environmental Research—UFZ
Article

DOI: 10.1007/s10533-011-9654-3

Cite this article as:
Docherty, K.M., Balser, T.C., Bohannan, B.J.M. et al. Biogeochemistry (2012) 109: 63. doi:10.1007/s10533-011-9654-3

Abstract

Wildfire in California annual grasslands is an important ecological disturbance and ecosystem control. Regional and global climate changes that affect aboveground biomass will alter fire-related nutrient loading and promote increased frequency and severity of fire in these systems. This can have long-term impacts on soil microbial dynamics and nutrient cycling, particularly in N-limited systems such as annual grasslands. We examined the effects of a low-severity fire on microbial biomass and specific microbial lipid biomarkers over 3 years following a fire at the Jasper Ridge Global Change Experiment. We also examined the impact of fire on the abundance of ammonia-oxidizing bacteria (AOB), specifically Nitrosospira Cluster 3a ammonia-oxidizers, and nitrification rates 9 months post-fire. Finally, we examined the interactive effects of fire and three other global change factors (N-deposition, precipitation and CO2) on plant biomass and soil microbial communities for three growing seasons after fire. Our results indicate that a low-severity fire is associated with earlier season primary productivity and higher soil-NH4+ concentrations in the first growing season following fire. Belowground productivity and total microbial biomass were not influenced by fire. Diagnostic microbial lipid biomarkers, including those for Gram-positive bacteria and Gram-negative bacteria, were reduced by fire 9- and 21-months post-fire, respectively. All effects of fire were indiscernible by 33-months post-fire, suggesting that above and belowground responses to fire do not persist in the long-term and that these grassland communities are resilient to fire disturbance. While N-deposition increased soil NH4+, and thus available NH3, AOB abundance, nitrification rates and Cluster 3a AOB, similar increases in NH3 in the fire plots did not affect AOB or nitrification. We hypothesize that this difference in response to N-addition involves a mediation of P-limitation as a result of fire, possibly enhanced by increased plant competition and arbuscular mycorrhizal fungi–plant associations after fire.

Keywords

FireGrasslandSoil microbiologyPLFAGlobal change

Introduction

Many grassland ecosystems, especially those in the semi-arid western United States, rely on fire disturbance regimes to maintain plant community structure and ecosystem function (Vogl 1979; Sugihara et al. 2007). Fires are an important factor in maintaining and restoring plant community composition and avoiding ecosystem degradation by invasive species (Menke 1992; DiTomaso et al. 2006). Fires may also indirectly improve ecosystem functions such as increasing soil organic matter and maintaining soil fertility (Menke 1992).

Fire regimes have been altered greatly by human activity, and are likely to be altered further in the coming century by climate change in the western US (Fried et al. 2004; Westerling et al. 2006; Westerling and Bryant 2008). Specifically, fires in California are likely to burn more rapidly and with higher intensity in the coming century under warmer climate scenarios (Fried et al. 2004). Multiple interacting effects of regional and global climate change may also increase fire incidence. For instance, increases in nitrogen deposition and fertilization, altered precipitation and drought patterns, and higher surface temperatures may substantially increase aboveground plant biomass, thus increasing fuel for fires (Henry et al. 2006).

Changes in aboveground communities as a result of fire can indirectly affect belowground systems by altering nutrient inputs, increasing solar warming and surface soil temperatures, and changing evapotranspiration rates and soil moisture availability (Neary et al. 1999). Furthermore, fire can lead to altered production of greenhouse gases from soils, such as N2O (Niboyet et al. 2011) or CO2 (Schulze et al. 2000; Dooley and Treseder 2011), suggesting that increased incidence of fire could alter current global change predictions (Schulze et al. 2000; Dooley and Treseder 2011). To date, few studies have linked the effects of grassland fire and other interacting global change factors to belowground microbial dynamics. Microbial communities, which comprise the majority of Earth’s biomass and diversity (Roesch et al. 2007; Schloss and Handelsmann 2006), are responsible for many important environmental processes (e.g. Kowalchuk and Stephen 2001; Hayatsu et al. 2008; Evans and Wallenstein 2011; Yavitt et al. 2011) and are central to the long-term response of ecosystems to global change (Hu et al. 1999; Singh et al. 2010; Brown et al. 2011, Todd-Brown et al. 2011, Treseder et al. 2011; Wallenstein and Hall 2011). Yet little is known about how wildfire affects microbial community structure and the subsequent ecosystem functions that microorganisms mediate.

Low-severity grassland fires alter both physical and chemical soil properties that affect microbial communities, particularly within the top 15-cm of soil where most aerobic microbial activity occurs (Raison 1979; Neary et al. 1999). In the short-term, direct heating can lead to microbial mortality in the soil (Raison 1979), though this effect is reduced in low-severity grassland fires where aboveground biomass (AGB) and litter are ashed quickly with little downward transfer of heat (Hart et al. 2005; Raison 1979). However, removal of AGB leads to increased soil warming and drying through direct solar irradiation (Raison 1979; Henry et al. 2006). This can have longer-term effects on plant productivity, microbial respiration, and decomposition rates (Fioretto et al. 2005; Hamman et al. 2007).

Chemically, low-severity fires deposit plant ash residues that contain macronutrients and organic carbon (Boerner et al. 2006). Specifically these include: carbonates of alkaline earth metals, silica, heavy metals, sesquioxides, phosphates, and organic and inorganic N (Raison 1979; Grogan et al. 2000; Wan et al. 2001; Picone et al. 2003; Li and Herbert 2004; Boerner et al. 2006). Additionally, heating from fire can volatilize nitrogen in the top few centimeters of the soil horizon (Wan et al. 2001). Particularly in low-severity summer grassland fires, which occur during the dry period after plants have died, this increase in nutrients through fire ash can have effects that emerge in the following growing season(s) (Henry et al. 2006).

Physical and chemical changes induced by fire can alter soil microbial dynamics in several ways. Depending on the severity of the fire, total soil microbial biomass can decrease dramatically immediately following the fire (Boerner et al. 2005). Generally, fires reduce fungal biomass, while bacterial biomass is unaltered (Dooley and Treseder 2011). The activity of microbial enzymes for organic carbon degradation often decreases after burning (Eivazi and Bayan 1996; Ajwa et al. 1999; Gutknecht et al. 2010). Similarly, activities of microbial enzymes involved in phosphorus metabolism can also decrease after burning (Ajwa et al. 1999; Gutknecht et al. 2010). However, the responses of microbial enzymes involved in organic nitrogen decomposition can vary after fire, with both increases (Ajwa et al. 1999) and decreases (Gutknecht et al. 2010) being reported. Few studies have examined the effects of fire on specific microbial functional groups linked to a particular biogeochemical rate. Phylogenetically distinct microorganisms show varied physiological responses to stress and disturbance, with different implications for microbial biogeochemical cycling (Schimel et al. 2007). Since fire can induce large changes, particularly in available N, responses of particular groups and sub-groups of microorganisms to a flush of available nutrients may result in long-term changes at an ecosystem level. For example, Yeager et al. (2005) show that high-severity forest fires can have multi-seasonal effects on the community structure of soil nitrifying bacteria and nitrification rates.

However, the mechanisms by which ecosystem changes conveyed by fire will affect soil microbial communities in combination with other global changes are almost completely unknown. Shifts in plant–microbe competition for novel resources deposited by fire can affect the timing of microbial responses to fire (Fiorietto et al. 2005; Hart et al. 2005; Gutknecht et al. 2010). However, shifts in competition can be further confounded by factors such as increased plant productivity as a result of nitrogen deposition, or suppressed plant productivity as a result of elevated atmospheric CO2 (Shaw et al. 2002; Zavaleta et al. 2003; Henry et al. 2006).

In this study, we examined multi-year responses to a low-severity grassland fire interacting with three other global change manipulations (nitrogen deposition, elevated CO2 and elevated precipitation) at the Jasper Ridge Global Change Experiment (JRGCE), near Palo Alto, CA. We examined microbial community responses for 3 years post-fire at several levels of resolution, including total microbial growth (lipid biomass), general microbial community structure (lipid composition), and the abundance and community structure of a bacterial gene involved in nitrification. Lipid analysis allows for the quantitative analysis of total microbial biomass and several ecologically distinct groups of microorganisms, including Gram-positive bacteria, Gram-negative bacteria and arbuscular mycorrhizal fungi (AMF) (Smithwick et al. 2005; Balser et al. 2005; Carney and Matson 2006; Gutknecht et al. 2011). Because fire has been shown to alter soil N inputs which can impact soil microbial N-transformations at JRGCE (Niboyet et al. 2011) and the JRGCE is traditionally N-limited (Henry et al. 2006), we also chose to examine ammonia-oxidizing bacteria (AOB) as a functional group that is likely impacted most quickly by both direct and indirect effects of fire. We compared microbial community parameters to multi-year responses in plant productivity and abiotic factors to fire in the context of other multi-factor global change manipulations. Additionally we examined the effects of fire on nitrification potential and AOB communities at one time point following the fire. Because fire can increase soil nutrient inputs, we predicted that the soil microbial community would respond to fire and nitrogen deposition treatments in a similar manner. We also predicted that the affects of fire would dampen over time as the soil community recovered. This study provides a whole-ecosystem context to the response of a California annual grassland to fire coupled with future global change scenarios.

Methods

Field site and experimental design

Soils were sampled for 4 years from the JRGCE from 2003 to 2006 (beginning 6 years after treatments began in 1998). The JRGCE is located in the eastern foothills of the Santa Cruz Mountains (37°40N, 122°22W) at the Jasper Ridge Biological Preserve (Stanford University). The region experiences a Mediterranean climate with a cool wet growing season and hot dry summer. The plant community at the site is typical of a Californian annual grassland, dominated by non-native annual grasses (Avena barbata, Avena fatua, Bromus hordeaceus, and Lolium multiflorum; Zavaleta et al. 2003; Henry et al. 2006). Nonnative forbs (Geranium dissectum, Erodium botrys, and Crpis vesicaria) and native annual forbs (Hemizonia congesta ssp. luzulifolia and Epilobium brachycarpum) are also present (Henry et al. 2006). Native grasses (Danthonia californica and Nasella pulchra) are dominant perennials, but comprise less than 10% of total biomass at the JRGCE (Henry et al. 2006). The JRGCE is located on a fine, mixed, thermic Typic Haploxeralf soil type, consisting of weathered alluvium from the Franciscan complex (Kashiwagi 1985).

The JRGCE includes four global change treatments, each at two levels (ambient and elevated), in a full factorial design. The treatments consisted of: CO2 (at ambient or elevated to 700 ppm); N-deposition (ambient or +7 g of N (as Ca(NO3)2) m−2); precipitation (ambient or 50% above ambient); and warming (ambient or increased to approximately +1°C in the grass canopy); Shaw et al. 2002; Zavaleta et al. 2003). The JRGCE has been under continuous manipulation of all four treatments since the beginning of the 1998–1999 growing season (November 1998). The treatment levels are consistent with regional and global predictions over the twenty first century (Hayhoe et al. 2004). In July 2003, an accidental fire burned a portion of the experiment. A controlled burn was then conducted such that two complete experimental blocks were burned, adding a fifth treatment to the experimental system. The burns resulting in the 2-block fire treatment were rapid, low severity fire, typical of wildfires in California annual grasslands that ashed above-ground litter in the plots. Warming treatments in the burn blocks were removed to create a smaller ‘burn experiment’ which had four replicates each of burning, elevated CO2, N-deposition, and precipitation, all fully crossed, for a total of 16 treatments. In order to test wildfire effects and interactive effects between wildfire and global change factors, we analyzed a subset of data including all plots in the two burn blocks (N = 32) and all un-warmed, unburned blocks under the other global change treatments (total N = 80 for our analysis).

Plant biomass and soil sampling

At the JRGCE there are two annual peaks (mid versus late spring) in plant AGB production that are dependent both on species and treatments (Zavleta et al. 2003; Dukes et al. 2005; Henry et al. 2006). Plant AGB data were analyzed from the first biomass peak on the following dates: 24 April, 2003 (4 months before the fire); 22–26 April, 2004 (9 months after fire); 25–26 April 2005 (21 months after fire), 9–11 May 2006 (33 months after fire). AGB data from the first biomass peak was thought to correlate more closely with our microbial analyses from soil samples taken on the same date. AGB was determined from the collection of aboveground plant matter from a 141 cm2 area of each experimental plot. Litter mass was determined by collecting all senesced plant material from within the same 141 cm2 area. Belowground biomass (BGB) was determined only at the first harvest date each year (the same dates as above) by separating live roots out of soil cores (15 cm depth) taken in the area of the first AGB harvest. All biomass was oven-dried (70°C) before weighing (Dukes et al. 2005). Directly after AGB was harvested, four replicate soil cores (2.5 cm wide and 15 cm deep) were taken from the same 141 cm2 area of each experimental plot and thoroughly homogenized. Sub-samples were then stored appropriately for each analysis described below.

Abiotic soil measurements

We measured the percent soil moisture gravimetrically every year from each freshly collected soil sample. We determined the percent water content from 10 g of soil, dried for 1 week at 105°C (Klute 1986). We measured soil N-NH4+ concentrations only in April 2004 with an Alpkem analyzer (Alpkem, Wilsonville, OR, USA). Ammonium extracts were prepared from 15 g fresh soil shaken in 1 M KCl and frozen at −20°C before analysis. We measured soil pH every year on each frozen soil sample. pH was determined from 5 g of thawed soil suspended in 10 ml deionized water, using an Accumet pH electrode and meter (Avrahami and Bohannan 2009). Soil temperature, soil moisture, pH and N-NH4+ concentrations were used to calculate molar concentrations of NH3 per gram of soil, which are reported throughout the manuscript as the available form to soil ammonia oxidizers (Avrahami and Bohannan 2009).

Microbial lipid analysis

We extracted lipids as described in detail in Gutknecht et al. (2011). In brief, we extracted fatty acids from 3 g of freeze-dried soil using a chloroform and phosphate buffer extraction, followed by reduction of the chloroform phase. We then performed saponification followed by strong acid methanolysis to extract the methyl-esterfied fatty acids. We identified lipid peaks using bacterial fatty acid standards and MIDI peak identification software (“Sherlock microbial identification system”, MIDI Inc, Newark, DE). Peak areas were converted to nmol g soil−1 using internal standards (9:0 nonanoic methyl ester and 19:0 nonadecanoic methyl ester). The total nmol lipid g dry soil−1 (sum of all lipids present, 20 or less carbon atoms in length) was used as an index of microbial biomass (Vestal and White 1989; Zelles et al. 1992; Frostegård and Bååth 1996). We used individual lipids as biomarkers of broad groups of the microbial community: 16:1 ω5c for AMF (Balser et al. 2005), 18:2 ω6,9c for saprotrophic fungi (SF; Balser et al. 2005), 16:1 ω7c for Gram-negative bacteria (Wilkinson et al. 2002), and 15:0 iso for Gram-positive bacteria (Wilkinson et al. 2002). The fungal to bacterial ratio was calculated as the average biomass of fungal lipid biomarkers (16:1 ω5c, 18:1 ω9c, and 18:2 ω6,9c) divided by the average biomass of bacterial lipid biomarkers (15:0 iso, 15:0 anteiso, 16:0 2OH, 16:0 iso, 16:1 ω7c; 16:0 10 methyl; 17:0 iso, 17:0 anteiso, 17:0 cyclo, 18:1 ω5c, 18:0 10 methyl, and 19:0 cyclo) as described by Frostegård and Bååth (1996).

Ammonia oxidizer analysis

DNA extraction

Samples from the first growing season after fire (April 2004) were also examined for ammonia-oxidizer functional genes and nitrification rates. Soils for DNA analysis were stored at −80°C after field collection until laboratory analyses could be performed. We extracted DNA from 0.5 g frozen soil using the Fast DNA® SPIN® Kit for Soil (BIO 101, Carlsbad, CA, USA) and the FastPrep Instrument (BIO 101, Carlsbad, CA, USA), according to the manufacturer’s instructions (Avrahami and Bohannan 2007). The DNA was re-suspended in a final volume of 50 μl and stored at −80°C.

PCR amplification of Bacterial amoA for T-RFLP analysis

We amplified the Bacterial amoA gene from DNA extracts according to Avrahami and Bohannan (2007). We used the modified forward primer amoA-1F* from Stephen et al. (1999) originally described in Rotthauwe et al. (1997). The 5′ end of this primer was labeled with 5-carboxyfluorescein for T-RFLP analysis. We used a degenerate reverse primer amoA-2R (Rotthauwe et al. 1997). Amplification was performed by using 1 μM of each primer (QIAGEN, Valencia, CA, USA), 0.48 μg μl−1 bovine serum albumin (Rosch, Penzberg, Germany), 0.625 units of AmpliTaq DNA polymerase (AmpliTaq, Applied Biosystems, Foster City, CA, USA), and 12.5 μl of MasterAmp 2′ PCR premix F containing 100 mM Tris–HCl (pH 8.3), 100 mM KCl, 7 mM MgCl2, 400 μM of each deoxynucleoside triphosphate, and PCR enhancer betaine (Epicentre Biotechnologies, Madison, WI, USA). We added DNA template and sterile water to a final volume of 25 μl. All reactions were performed in triplicate. We started amplifications by placing PCR tubes into a preheated (94°C) DNA Engine thermal cycler (MJ Research, San Francisco, CA, USA). The amplification program included 5 min at 94°C, followed by 35 cycles of 45 s at 94°C, 1 min at 55°C, and 1 min at 72°C, and 10 min at 72°C for the last cycle. Following a gel electrophoresis check, the three replicates were pooled and cleaned together using a Wizard DNA Clean-Up Kit (Promega, Madison, WI, USA).

AOB community structure

We examined Bacterial amoA PCR products using terminal restriction fragment length polymorphism (T-RFLP) analysis. PCR products were restricted with Bsh1236I (Fermentas, Hanover, MD, USA), according to manufacturer’s instructions (Horz et al. 2004; Avrahami and Bohannan 2007). This enzyme cuts at the sequence CG/CG, and produces a 434 bp product which is specifically diagnostic for phylogenetic Cluster 3a AOB most closely related to Nitrosospira multiformis (Horz et al. 2004; Avrahami and Bohannan 2007). Total peaks produced using this T-RFLP procedure ranged from 35 to 490 bp in size. We called peaks within an error of +5 and −5 bp in size as diagnostic for the same phylogenetic cluster. Thus, peaks ranging from 429 to 439 bp in size were considered to be in the same group as peaks that were 434 bp in size, and diagnostic for Cluster 3a ammonia-oxidizers. The relative proportion of the 434 bp-sized peak represents a metric for examining a shift in AOB community structure, corresponding to an increase or decrease in the relative abundance of this specific phylogenetic Cluster 3a within the whole AOB community.

We performed T-RFLP fingerprinting analysis at the Genomics Technology Support Facility at Michigan State University (East Lansing, MI, USA), as described in Horz et al. (2004). The relative abundances of individual restriction fragments (or T-RFs) in a given amoA PCR product were calculated based on the peak height of the individual T-RFs in relation to the total peak height of all T-RFs detected in the respective T-RFLP community fingerprint pattern. The peak heights were automatically quantified by the GENESCAN software (Perkin-Elmer, Foster City, CA, USA).

Real-time PCR amplification of Bacterial amoA

For the DNA extracts, we also used a nested real-time quantitative PCR approach (Q-PCR) as described in Avrahami and Bohannan (2007) to determine the total abundance of Bacterial amoA in the test soils. In brief, the nested approach was performed with both rounds of amplification using the same primer set as described above for end-point PCR without the addition of a fluorescent label to the 5′ end of the forward primer (Stephen et al. 1999, modified from Rottahuwe et al. 1997). Thus, this method provides a measure of total abundance of all AOB, while the T-RFLP method provides a measure of relative abundance Cluster 3a AOBs only.

Nitrification potential measurements

Nitrification potential reflects the enzymatic potential of the soil nitrifying bacteria to oxidize NH3 into NO2 or NO3 under optimal conditions (Barnard et al. 2005). For samples collected in April 2004, nitrifying enzyme activity was measured as described in Avrahami and Bohannan (2007). Sterile Erlenmeyer flasks containing 45 ml phosphate buffer (1 mM; pH 7.4), 0.04 ml (NH4)2SO4 (0.25 M), and 5 g of soil were incubated at 25°C on a shaker for 46 h. Samples were taken at five time points, centrifuged for 5 min, and filtered through Acrodisc-CR 25-mm syringe filters (0.2 μm; Life Sciences). Samples were stored at −20°C until analysis of nitrite and nitrate in a DIONEX ion chromatograph system (Sunnyvale, CA) equipped with an AS11-HC analytical column (Avrahami and Bohannan 2007). The sum of NO2 and NO3 concentrations increased linearly with time. Rates of potential nitrification activity were determined from the slope of a linear regression of nitrite plus nitrate concentration on time (Avrahami and Bohannan 2007).

Statistics

We used Nonmetric Multidimensional Scaling (NMS), a multivariate analysis technique, to fingerprint the microbial community using the R software vegan package for multivariate analysis (R core team 2010). NMS is a useful method because it does not require data normality or equal variances, and also does not produce the artifacts of other ordination methods such as principal components analysis (McCune and Grace, 2002). Lipids present from all years of the study in greater than 0.005 mol fraction (moles individual lipid−1 × moles total lipid biomass−1) were used to provide a breadth of information while also preventing “noise”. Sørenson (Bray) distance was used to construct distance matrices for analysis. Thirty independent runs were performed with raw data to ensure that the best solution is global and not reflecting only local minima (McCune and Grace 2002; R core team 2010).

We performed Permanova analysis using R statistical software to test for treatment effects on the multivariate PLFA-based microbial community (R core team 2010; Anderson 2001). The analysis we used was based on the JRGCE split plot, fully crossed design. The dependent variables elevated CO2, elevated N-deposition, warming, elevated precipitation, and all possible interactions were used on 2003 data to test main and interactive treatment effects. The dependent variables elevated CO2, elevated N-deposition, fire, elevated precipitation, and all possible interactions were used on 2004–2006 data to test main and interactive treatment effects. As with NMS analysis, we used Sørenson (Bray) distance to construct the distance matrix for analysis. We performed 1000 permutations for the best possible sensitivity of calculated F-statistics and subsequent p values.

We used Analysis of Variance (ANOVA) on individual microbial indicators to complement the multivariate analysis. The dependent variables analyzed include total lipid biomass, the fungal to bacterial ratio, Gram-negative, Gram-positive, AMF and general fungal lipids biomarkers, the proportion of AOB Cluster3a (T-RFLP), AOB abundance (Q-PCR), Nitrification Potential, AGB, BGB, litter, pH, and NH3 data for all years these data were available. ANOVA analysis was performed using PROC MIXED in SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) based on the fully blocked, split-plot design (factors nitrogen and water assigned randomly at the subplot level, nested within CO2 and warming treatments assigned at the plot level) of the JRGCE including independent variables elevated CO2, nitrogen addition, fire, water addition, and all possible interactions (in 2003 warming was included instead of fire). For 2003 data, N = 128, and in 2004, 2005 and 2006, N = 80 to reflect a balanced design to test for the effect of fire on these parameters. The only exception was that for April 2006, we only measured soils that were not influenced by the fire for [NH3]. Thus, in April 2006, N = 96, and we were unable to assess the effect of fire on soil [NH3] for this particular year. Means were calculated as least square means and the degrees of freedom were estimated using the Satterthwaite approximation. Transformations for all data for split-plot ANOVAs were as listed in the previous NMS section.

Results

Fire effects on plant biomass and abiotic measurements

We focused on biomass production at the first spring biomass peak measured at the same time as our microbial analysis [see Henry et al. (2006) for a detailed discussion of maximum biomass after fire]. The wildfire imposed at the JRGCE in July 2003 significantly altered AGB and litter quantity in the April 2004 sampling (Fig. 1; Table 1). In April 2004, 9 months after the fire, we found that the unburned (or ambient) plots had an average AGB of 281 g m−2 (±1.1) during the first harvest, while the burned plots were significantly higher, up to 508 g m−2 (±1.1). This trend was reversed in April 2005, 21 months after the fire, when the average AGB at the first harvest was significantly lower in the burned plots [382 g m−2 (±1.1)] than in the unburned plots [511 g m−2 (±1.1)] (Figs. 1, 2b; Table 1). By May 2006, 33 months after the fire, AGB in previously burned plots and ambient plots were indistinguishable (Table 1, Supplementary Table 1). Conversely, we found that average litter quantity was significantly lower in the burned plots during the April 2004 sampling, with 195 g m−2 (±1.4) litter in the unburned plots and only 8.3 g m−2 (±1.6) in the burned plots (Fig. 1; Table 1). This trend was still present in April 2005, though the difference between the burned and unburned plots was not statistically significant (Fig. 1; Table 1). By May 2006, there was no difference in litter quantity between the previously burned plots and the ambient plots (Table 1). BGB was not altered by the fire treatment alone in any of the years following the burn (Table 1). In April 2004, the only year we measured N-NH4+, we found that calculated NH3 concentrations were significantly higher in the burned plots than in unburned plots and that pH was not affected by fire (Table 1; Fig. 2a). Fire also has been shown to lead to long-term increases in soil moisture in these plots, as described in Niboyet et al. (2011).
https://static-content.springer.com/image/art%3A10.1007%2Fs10533-011-9654-3/MediaObjects/10533_2011_9654_Fig1_HTML.gif
Fig. 1

Effects of fire on plant productivity (top) and microbial lipid composition (bottom) at JRGCE. Data collected in April 2003 were pre-fire. The fire occurred in July 2003. Data collected in April 2004 were 9 months post-fire; data collected in April 2005 were 21 months post fire. Stars reflect the level of significance between the least squared means derived from the split-plot ANOVA model. Error bars represent one standard error associated with the statistical least squared means model. Aboveground biomass (AGB) and litter were significantly different in burned (white) versus unburned (black) plots in April 2004 (AGB: F1,11 = 12.25, p = 0.005; litter: F1,11 = 28.40, p = 0.0002). AGB was significantly different in burned (white) versus unburned (black) plots in April 2005 (F1,11 = 9.59, p = 0.003). There were no significant differences in belowground biomass (BGB). Gram-positive bacterial lipid relative abundance (black, mol%) was significantly different in burned versus unburned plots in April 2004 (F1,10.4 = 5.55, p = 0.039) and Gram-negative bacterial lipid relative abundance (dark gray, mol%) was significantly different in burned versus unburned plots in April 2005 (F1,10 = 11.49, p = 0.007)

Table 1

Summary of p values from split-plot ANOVAs to test the treatment effects of fire and global changes on the following factors: above-ground biomass = log g m−2; below-ground biomass = log g m−2; litter = log g m−2; pH = pH in water; soil moisture = gravimetric soil moisture (square root arcsin transformed %); NH3 (2004 only) = calculated ammonia from extractable N-NH4+ (log M)

 

Above-ground biomass

Below-ground biomass

Litter

pH

NH3

April 2003 (pre-fire)

 CO2

0.246

0.343

0.718

0.996

n.a.

 Warming

0.406

0.962

0.748

0.507

n.a.

 Precipitation

0.049

0.298

0.633

<0.0001

n.a.

 N-deposition

<0.0003

0.873

0.074

<0.0001

n.a.

April 2004 (9 months post-fire)

 CO2

0.947

0.012

0.934

0.698

0.247

 Precipitation

0.358

0.728

0.383

<0.0004

0.973

 N-deposition

0.001

0.263

0.495

0.587

<0.0001

 Fire

0.005

0.993

<0.0003

0.253

0.008

 Fire × CO2

0.015

0.042

0.151

0.585

0.006

 Fire × Precipitation

0.296

0.799

0.961

0.450

0.416

 Fire × N-deposition

0.100

0.154

0.382

0.608

0.001

 Fire × CO2 × N-deposition

0.889

0.980

0.352

0.590

0.860

 Fire × CO2 × Precipitation

0.937

0.657

0.751

0.998

0.418

 Fire × N-deposition × Precipitation

0.197

0.800

0.431

0.108

0.315

April 2005 (21 months post-fire)

 CO2

0.001

0.070

0.003

n.a.

n.a.

 Precipitation

0.904

0.115

0.915

n.a.

n.a.

 N-deposition

<0.0002

0.731

0.041

n.a.

n.a.

 Fire

0.003

0.336

0.093

n.a.

n.a.

 Fire × CO2

0.620

0.264

0.620

n.a.

n.a.

 Fire × Precipitation

0.339

0.005

0.406

n.a.

n.a.

 Fire × N-deposition

0.508

0.243

0.606

n.a.

n.a.

 Fire × CO2 × N-deposition

0.493

0.913

0.706

n.a.

n.a.

 Fire × CO2 × Precipitation

0.440

0.016

0.752

n.a.

n.a.

 Fire × N-deposition × Precipitation

0.478

0.997

0.496

n.a.

n.a.

May 2006 (33 months post-fire)

 CO2

0.675

0.654

0.021

n.a.

n.a.

 Precipitation

0.265

0.451

0.145

n.a.

n.a.

 N-deposition

0.148

0.492

0.005

n.a.

n.a.

 Fire

0.285

0.898

0.095

n.a.

n.a.

 Fire × CO2

0.495

0.434

0.246

n.a.

n.a.

 Fire × Precipitation

0.815

0.069

0.781

n.a.

n.a.

 Fire × N-deposition

0.643

0.476

0.004

n.a.

n.a.

 Fire × CO2 × N-deposition

0.628

0.875

0.891

n.a.

n.a.

 Fire × CO2 × Precipitation

0.548

0.758

0.934

n.a.

n.a.

 Fire × N-deposition × Precipitation

0.801

0.582

0.072

n.a.

n.a.

2003 values were based on all 128 plots of the JRGCE, while 2004 and 2005 data were based on the burn subset of data (see “Methods”). Designation of n.a. indicates that consistent data were not available. While all possible treatment interactions were tested, only treatments involving fire are shown (see Supplementary Table 1 for full information). Bold numbers indicate significance at p < 0.05

https://static-content.springer.com/image/art%3A10.1007%2Fs10533-011-9654-3/MediaObjects/10533_2011_9654_Fig2_HTML.gif
Fig. 2

Compared effects of fire and N-deposition 8-months post-fire. Least squared means from split-plot ANOVA tests are presented for a soil ammonia (NH3) concentration, b aboveground biomass (AGB), c total ammonia-oxidizing bacterial (AOB) abundance as measured by quantitative PCR, d relative abundance of AOB Nitrosospira-like Cluster 3a, as measured by T-RFLP and e arbuscular mycorrhizal fungal (AMF) lipid relative abundance. White bars indicate either ambient N-deposition (left) or unburned (right); black bars indicate either elevated N-deposition (left) or burned (right). Error bars represent one standard error for the error associated with the statistical least squared means model. Stars indicate level of significance. NH3 concentration increased significantly in both elevated N-deposition (F1,48 = 21.77, p < 0.0001) and burned (F1,11 = 10.4, p = 0.008) plots. AGB also increased significantly in both elevated N-deposition (F1,48 = 11.40, p = 0.002) and burned (F1,11 = 12.25, p = 0.005) plots. AOB abundance (F1,48.8 = 6.11, p = 0.017) and Cluster 3a abundance (F1,43.3 = 33.95, p < 0.0001) increased in elevated N-deposition plots only

Other global change and interactive effects on plant biomass and abiotic measurements

While we saw many responses to global change that were only present sporadically (Table 1), we did not focus on how global change responses varied over time. A detailed examination of inter-annual variation at the JRGCE has been reported elsewhere on a variety of measurements (Shaw et al. 2002; Zavaleta et al. 2003; Barnard et al. 2005; Dukes et al. 2005; Gutknecht et al. 2010, 2011).

Increased N-deposition consistently increased plant AGB in all years of this study except for in May 2006, where it had no significant effect on AGB (Fig. 2b; Table 1, Supplementary Table 1). We also found that the elevated N-deposition treatment was related an increase in soil NH4+ concentrations compared to ambient plots in 2004 (Fig. 2a; Table 1). We did not find that N-deposition affected BGB or litter quantity significantly (Table 1). In April 2004, we found that an interaction between the fire and CO2 treatments consistently affected several variables. AGB responded to an interactive fire by CO2 effect, such that the combination of the fire and CO2 treatments increased AGB beyond the increase exhibited by the fire treatment alone (Table 1, Supplementary Table 1). The BGB also responded to the fire by CO2 interactive effect in April 2004, such that the fire and elevated CO2 combined reduced the suppressive effect of elevated CO2 alone on BGB (Table 1, Supplementary Table 1). Finally, the fire by CO2 interactive effect in April 2004 influenced soil NH4+ concentrations, such that the combined two treatments negated a decreased calculated soil NH3 concentration under CO2 alone (Table 1, Supplementary Table 1).

Fire effects on microbial biomass, the fungal to bacterial ratio, and lipid composition

We used several analyses to examine the effects of fire on soil microbial lipids, including multivariate community structure (Permanova analysis) and univariate measures of total microbial biomass (using the sum of all lipids present, 20 or less carbon atoms in length), the fungal to bacterial ratio, and analysis of several lipids diagnostic of broad microbial groups. In April 2004 (9 months after fire), our results indicate that fire affected the multivariate lipid composition (Table 2), as well as the Gram-positive bacterial lipid biomarker (15:0 iso). Specifically there was a lower relative abundance of Gram-positive lipid biomarkers in burned (1.74 mol% ± 0.06) than in unburned plots (2.01 mol% ± 0.08) (Fig. 1; Table 2). The AMF biomarker (16:1 ω5c) was also 1.94 mol% higher in burned than unburned, ambient plots in April 2004, but this effect was not significant (Table 2, Supplementary Table 2; Figs. 1, 2e). In April 2005 (21 months after the fire) the only lipid based parameter affected by fire was the Gram-negative bacterial lipid biomarker (16:1 ω7c), which was significantly lower in the previously burned plots (9.40 mol% ± 0.50) than in unburned plots (11.47 mol% ± 0.39) (Fig. 1; Table 2). By May 2006 (33 months after fire), no parameter based on microbial lipid analysis exhibited a response to fire (Table 2, Supplementary Table 2).
Table 2

Summary of p values from split-plot ANOVAs to test the treatment effects of fire and global changes on microbial indicators

 

Multivariate

Total microbial biomass

Fungal:bacterial lipids

General fungi

Gram-positive bacteria

Gram-negative bacteria

Arbuscular mycorrhizal fungi

Nitrification potential

April 2003 (pre-fire)

 CO2

0.667

0.345

0.138

0.354

0.492

0.075

0.615

n.a.

 Warming

0.1045

0.381

0.651

0.336

0.774

0.764

0.447

n.a.

 Precipitation

0.01

0.914

0.017

0.517

0.582

0.830

0.053

n.a.

 N-deposition

0.01

0.390

0.004

0.589

0.872

0.258

0.057

n.a.

April 2004 (9 months post-fire)

 CO2

0.463

0.628

0.287

0.132

0.372

0.604

0.346

0.946

 Precipitation

0.576

0.294

0.692

0.631

0.936

0.680

0.430

0.318

 N-deposition

0.025

0.092

0.029

0.042

0.077

0.002

0.112

0.005

 Fire

0.001

0.443

0.276

0.452

0.039

0.438

0.243

0.792

 Fire × CO2

0.323

0.762

0.157

0.388

0.233

0.403

0.195

0.334

 Fire × Precipitation

0.922

0.132

0.545

0.418

0.811

0.326

0.619

0.678

 Fire × N-deposition

0.418

0.057

0.726

0.009

0.010

0.225

0.721

0.072

 Fire × CO2 × N-deposition

0.737

0.642

0.969

0.814

0.884

0.595

0.811

0.135

 Fire × CO2 × Precipitation

0.441

0.976

0.031

0.264

0.367

0.236

0.028

0.235

 Fire × N-deposition × Precipitation

0.925

0.621

0.502

0.648

0.877

0.867

0.977

0.347

April 2005 (21 months post-fire)

 CO2

0.003

0.240

0.285

0.278

0.020

0.006

0.805

n.a.

 Precipitation

0.829

0.157

0.539

0.006

0.698

0.742

0.642

n.a.

 N-deposition

0.002

0.913

<0.0001

0.235

0.178

0.010

<0.0001

n.a.

 Fire

0.346

0.697

0.525

0.510

0.674

0.007

0.965

n.a.

 Fire × CO2

0.495

0.388

0.804

0.605

0.700

0.999

0.829

n.a.

 Fire × Precipitation

0.597

0.684

0.057

0.272

0.648

0.577

0.033

n.a.

 Fire × N-deposition

0.736

0.767

0.359

0.325

0.805

0.046

0.919

n.a.

 Fire × CO2 × N-deposition

0.969

0.262

0.414

0.035

0.566

0.406

0.743

n.a.

 Fire × CO2 × Precipitation

0.770

0.017

0.109

0.667

0.252

0.417

0.139

n.a.

 Fire × N-deposition × Precipitation

0.829

0.653

0.470

0.388

0.846

0.444

0.194

n.a.

May 2006 (33 months post-fire)

 CO2

0.130

0.042

0.482

0.100

0.749

0.003

0.101

n.a.

 Precipitation

0.701

0.396

0.857

0.060

0.542

0.308

0.939

n.a.

 N-deposition

0.732

0.502

0.003

0.377

0.770

0.980

0.131

n.a.

 Fire

0.114

0.488

0.893

0.350

0.857

0.755

0.644

n.a.

 Fire × CO2

0.042

0.304

0.197

0.697

0.123

0.114

0.758

n.a.

 Fire × Precipitation

0.374

0.496

0.928

0.353

0.580

0.145

0.923

n.a.

 Fire × N-deposition

0.860

0.848

0.726

0.690

0.573

0.821

0.575

n.a.

 Fire × CO2 × N-deposition

0.040

0.009

0.301

0.062

0.216

0.026

0.082

n.a.

 Fire × CO2 × Precipitation

0.546

0.452

0.409

0.998

0.842

0.343

0.995

n.a.

 Fire × N-deposition × Precipitation

0.715

0.713

0.973

0.740

0.888

0.549

0.456

n.a.

2003 values were based on all 128 plots of the JRGCE, while 2004–2006 data were based on the burn subset of data (see “Methods”). While all possible treatment interactions were tested, only treatments involving fire are shown (see Supplementary Table 2 for full information). Microbial indicators: Total microbial biomass (nmol lipid g soil−1); fungal:bacterial lipids = ratio of fungal to bacterial lipids; general fungi = 18:2 ω6,9c mol%, Gram-positive bacteria = 15:0 iso mol%; Gram-negative bacteria = 16:1 ω7c mol%; arbuscular mycorrhizal fungi = 16:1 ω5c mol%; nitrification potential in 2004 only (ng N h−1 g dry soil−1). The ‘multivariate’ column contains p values from Permanova nonparametric testing to determine treatment effects multivariately. Permanova analysis was performed on lipid relative abundance (mol fraction) data. The statistical model was based on the split-plot full factorial design of the JRGCE, run with 1000 permutations. Bold numbers indicate significance at p < 0.05

Other global change and interactive effects on microbial biomass, the fungal to bacterial ratio, and lipid composition

No consistent global change factor affected multivariate lipid composition, total microbial biomass, or any individual lipid biomarker in every year examined (Table 2, Supplementary Table 2). One very consistent result was that the fungal to bacterial lipid ratio was significantly lower in N-deposition plots in all 4 years examined in this study (Fig. 3b). In April 2003 we found a significant effect of elevated precipitation and N-deposition on multivariate lipid composition (Table 2, separate multivariate clustering between treated and untreated plots). Gram-negative bacterial biomarker relative abundance was significantly higher in elevated N-deposition plots in 2004 and 2005. In April 2004 the Gram-negative bacterial lipid biomarker increased from 9.57 mol% (±0.36) in ambient to 10.87 mol% (±0.37) under N addition; in April 2005 it increased from 9.79 mol% (±0.39) in ambient to 11.08 mol% (±0.41) under N addition, suggesting a consistent increase in the Gram-negative bacterial community as a result of N-deposition. In 2004 the general fungal lipid indicator increased from 2.75 mol% (±0.11) in ambient plots to 3.06 mol% (±0.12) in N-deposition plots. In 2005 there were significant effects of CO2 and N-deposition on multivariate lipid composition (Table 2). Additionally, AMF biomarker relative abundance was generally lower under N-deposition but this trend was only significant in April 2005, where relative abundance decreased from 17.89 mol% (±0.74) in ambient plots to 13.95 mol% (±0.75) in N-deposition plots (Table 2; Fig. 2e).
https://static-content.springer.com/image/art%3A10.1007%2Fs10533-011-9654-3/MediaObjects/10533_2011_9654_Fig3_HTML.gif
Fig. 3

Effects of fire and N-deposition on the ratio of fungal to bacterial lipids. Least squared means from split-plot ANOVA tests are presented for a burned (white) and unburned (black) plots and b ambient (white) and elevated (black) N-deposition treatments from April 2003, April 2004, April 2005, and May 2006. Error bars represent one standard error for the error associated with the statistical least squared means model. Stars indicate level of significance. The fungal to bacterial lipid ratio was significantly lower in all elevated N-deposition plots (2003: F1,50.5 = 9.17, p = 0.004; 2004: F1,44.1 = 5.12, p = 0.029; 2005: F1,40.3 = 30.62, p < 0.0001; 2006: F1,42.2 = 10.8, p = 0.003)

In addition to these single factors, we found a consistent interactive effect between the fire and N-deposition in the bacterial portion of the soil microbial community. In April 2004 (9 months after burning) elevated N-deposition treatments interacted such that when combined, the incidence of fire decreased the stimulatory effect of N-deposition on the Gram-positive bacterial lipid biomarker relative abundance (Supplementary Fig. 1a). Thus, under ambient levels of N-deposition, fire had no effect on Gram-positive bacterial lipids in April 2004, but under elevated N-deposition, fire decreased the Gram-positive bacterial lipid biomarker from 2.17 mol% (±0.08) to 1.70 mol% (±0.10) (Supplementary Fig. 1a). In April 2005 (21 months after burning) we found a similar fire by N-deposition interactive effect, except in the Gram-negative bacterial lipid biomarker instead of the Gram-positive bacterial biomarker (Supplementary Fig. 1b). Thus under ambient levels of N-deposition, fire had no effect on the Gram-negative bacterial lipid biomarker, but under elevated N-deposition, fire decreased the Gram-negative lipid biomarker from 12.61 mol% (±0.53) to 9.55 mol% (±0.60) (Supplementary Fig. 1b). By May 2006 (33 months after burning) this interactive effect of fire by N-deposition was no longer significant in any of the microbial lipids we measured (Table 2, Supplementary Table 2).

N-deposition and fire effects on a specific microbial functional group (ammonia-oxidizing bacteria) and rate (nitrification potential)

Our abiotic results show that both N-deposition and fire can significantly increase calculated NH3 concentrations in soils (Fig. 2; Table 1). Since NH3 is the direct substrate for the process of ammonia-oxidation and a rate-limiting step in nitrification, we chose to examine the similarities between these two global change factors further on the specific group of AOB and the process of nitrification in the first growing season following the fire (April 2004). We saw a response in both AOB and potential nitrification to nitrogen addition. The total AOB abundance (Fig. 2c) and the relative abundance of AOB Cluster 3a (Fig. 2d) were significantly higher in N-deposition plots than in the ambient (Fig. 2a). Nitrification potential measurements were also higher under N-deposition in April 2004. Nitrification potentials increased from 17.43 (±0.84) ng N h−1 g dry soil−1 in ambient plots to 24.4 (±0.79) ng N h−1 g dry soil−1 in N-deposition plots. No other global change factors or interactions significantly affected AOB abundance or Cluster 3a relative abundance. No other global change factors or interactions significantly affected AOB abundance or Cluster 3a relative abundance.

On the contrary, even though both the fire and nitrogen addition plots had similarly increased levels of calculated soil NH3, we found that after one growing season fire had no effect on AOB abundance (or quantity of ammonia-monooxygenase functional gene marker amoA, Fig. 2c) or community structure (relative abundance of Cluster 3a ammonia-oxidizers, Fig. 2d). We also found no effect of fire on nitrification potential measurements in April 2004 (Table 2). Nitrification potentials were approximately 475 ng N h−1 g−1 dry soil in both the ambient and burned plots.

Discussion

It should be noted that this is among the few studies to examine the response of microbial communities to grassland fires. Far more studies have focused on forest ecosystems (Sugihara et al. 2007; Dooley and Treseder 2011). Major differences between forest and grassland fires include the severity and heat of the fire, as well as the amount of resources deposited to the system as ash (e.g. Dooley and Treseder 2011; Hamman et al. 2007; Sugihara et al. 2007; Neary et al. 1999; Raison 1979). Additionally, forest and grassland ecosystems differ hugely in terms of soil formation, soil chemistry, and nutrient availability, which can directly affect microbial biodiversity (Brady and Weil 2008; Antonopoulos et al. in preparation). Forest microbial communities may also rely more on fermentation and organic acid pathways while grassland communities rely more on sugar alcohol and PO4 metabolism pathways (Antonopoulos et al. in preparation). Thus, we focus our discussion mainly on responses to fire in grassland ecosystems.

Aboveground biomass was altered for 2 years after a summer wildfire at the JRGCE

Our results indicate that aboveground plant productivity (AGB) was stimulated 9 months following fire in April 2004 (Figs. 1, 2b). Such increases in primary production after fire are not typically observed in Mediterranean-climate annual grassland systems, such as this one, where litter accumulation is low and fires are typically of low intensity and severity (Launchbaugh 1964; Hurlbert 1969) but are more typical of tallgrass prairie systems (Abrams et al. 1986; Briggs and Knapp 1995). Even at the JRGCE, when plant biomass was analyzed over the entire growing season (as opposed to the first peak in April 2004), no differences were observed between burned and unburned plots (Henry et al. 2006).

We suggest that the stimulatory effect of fire we observed on first peak AGB in the April 2004 harvest is the result of several indirect effects of fire on the ecosystem. First, decreased litter in the burned plots allowed for earlier germination and plant growth. There is evidence for this from the JRGCE normalized difference vegetation index data measured through the 2003–2004 growing season after the burn, which showed earlier canopy greenness in the burned plots than the unburned plots (Henry et al. 2006), despite a decrease in soil moisture in burned plots also observed at this time (Niboyet et al. 2011). This response of earlier season, but not maximum, plant growth could be why there was no significant difference in litter mass the following year (April 2005). Second, calculated NH3 concentrations significantly increased in the burned plots in April 2004, providing labile N to stimulate plant growth. The increases we and others have observed in AGB as a result of Ca(NO3)2 addition indicate that the Jasper Ridge system is N limited (Table 1; Shaw et al. 2002; Dukes et al. 2005; Menge and Field 2007). Productivity in N-limited systems with moderate fire frequencies can increase because of this functional N-fertilization effect of inorganic N-containing ash (Rissler and Parton 1982; Hurlbert 1988; Neary et al. 1999; Romanyá et al. 2001; Henry et al. 2006), though this is not always the case (Ajwa et al. 1999). In our study, increases in calculated soil NH3 in April 2004 can be compared to a similar study where levels of inorganic N (NH3 and NO3) nearly doubled in burned grassland soils (Picone et al. 2003).

Henry et al. (2006) suggested that in addition to alleviating N-limitation, fire can alleviate limitation by other mineral nutrients such as phosphorus (P). This hypothesis was formed from the observation that an increase in available PO4 ions after fire may remove the suppressive effect that elevated CO2 has on primary productivity at JRGCE (Henry et al. 2006; Shaw et al. 2002). In a similar study, available P increased significantly after burning as a result of the release of phosphates by heating, as well as calcium polyphosphates in ash (Picone et al. 2003). This explanation is corroborated by the observation that annual grasses in the burned plots in April 2004 had lower N:P ratios than in unburned plots, coupled with a shift in plant P-allocation to AGB (Henry et al. 2006). While it was beyond the scope of the present study to directly measure soil P-availability, it is a reasonable hypothesis that more phosphorus becomes available following a grassland burn, and that P-limitation may become alleviated in the Jasper Ridge system as a result of this added phosphorus.

In April 2005, 21 months after the fire, primary productivity was significantly less in burned than in unburned plots. However, microbial extracellular enzyme activities per unit microbial biomass were lower in previously burned plots than in unburned plots, suggesting that the microbial community was less nutrient limited at that time (Gutknecht et al. 2010). It is possible that certain nutrients, especially micronutrients, were more limiting for plants than microbes 21 months after fire, though further investigation is necessary to validate this hypothesis. Alternatively, although we only saw a marginally significant increase in litter in burned plots 21 months after fire, it is possible that this litter or standing biomass from the previous season inhibited germination or delayed early season growth in the 2004–2005 growing season. By May 2006 there were no effects of fire on AGB, and microbial extracellular enzyme activities were similar to those in ambient plots (Gutknecht et al. 2010). This suggests that by 33 months post-fire, whole-ecosystem productivity had recovered completely from wildfire disturbance.

Microbial community structure, but not total microbial biomass, was affected for 2 years following a grassland fire

Total microbial biomass in our study was not affected by fire in any year (Table 2). In their meta-analysis of the effects of fire on microbial biomass presented in this special issue, Dooley and Treseder (2011) show that on average, microbial biomass tends to decline significantly following a wildfire, but that this decrease is often related to fire severity. In less severe fires, such as prescribed burns, that are similar to the JRGCE fire in July 2003, microbial biomass is not decreased by fire in any type of biome (Dooley and Treseder 2011). Several studies have shown that after a low-intensity grassland fire, soil chemical properties change (Sherman et al. 2005) but that total soil microbial biomass does not change long-term (Ponder et al. 2009; Hamman et al. 2007; Song et al. 2004; Picone et al. 2003).

While total microbial biomass was not altered in our study, multivariate lipid composition was affected by fire in April 2004, 9 months after fire, but only at this time point (Table 2). At a finer scale, our results revealed responses to fire in particular microbial groups both 9 months (April 2004) and 21 months (April 2005) following the fire (Fig. 1). All effects of the fire were then absent in the microbial community by 33 months following the fire (May 2006, Supplementary Table 2).

In April 2004, 9 months after fire, Gram-positive bacterial relative abundance was lower in burned plots, with similar non-significant trends in the Gram-negative bacteria and general fungi (not including AMF), while the AMF indicator relative abundance was not different in burned plots. We did see a marginal increase in the fungal to bacterial lipid ratio (Fig. 3; Table 2) in April 2004, which was due to a non-significant increase in the AMF indicator relative abundance. Several studies have shown the opposite, that within the year following a low-severity fire, soil fungal biomass decreases (Hamman et al. 2007; D’Ascoli et al. 2005; Ponder et al. 2009; Dooley and Treseder 2011), while bacteria recover rapidly and are more resilient to (D’Ascoli et al. 2005) or even stimulated by (Ponder et al. 2009) fire. Our results may differ from these other studies because most comparable studies do not distinguish between general fungal indicators (18:2 ω6,9c, 18:3 ω6c, 18:ω9c; Hamman et al. 2007; Ponder et al. 2009) from arbuscular mychorrhizal fungal indicators (16:1 ω5c; Balser et al. 2005), as we did. In studies that examined mycorrhizal biomass specifically, AMF abundance increased in low-frequency burns in African savannah systems (Hartnett et al. 2004), but no other studies have directly assessed the responses of AMF to grassland fire. In the next section we discuss in more detail the possible implications of these shifts in microbial and fungal communities in the first growing season post-fire.

In April 2005, 21 months post-fire, only the Gram-negative bacterial indicator was still affected by fire, being significantly lower in burned than in unburned plots, while other lipid indicators were unaffected by fire (Fig. 1). Gram-negative bacteria are the predominant rhizosphere bacteria, and are dependent on labile carbon and nutrients excreted from roots. Although we saw no change in BGB in 2005 in burned plots, the lower aboveground plant biomass we observed may have corresponded to a decreased amount of labile carbon flow to the rhizosphere, thus limiting nutrients and growth to Gram-negative bacteria.

Specific nitrifying microorganisms are affected by nitrogen addition but not fire in the Jasper Ridge system

While lipid analysis is a good method for detecting proportional changes in bacterial and fungal communities, a drawback is that it cannot provide information about specific phylogenetic groups. Since N-deposition and fire both resulted in similar increases in AGB and calculated soil NH3 (Fig. 2a) in the growing season immediately following fire (April 2004), we complemented the lipid analysis for that time point with a finer-scale examination of the ammonia-oxidizing bacterial (AOB) community. Through the ammonia-monooxygenase enzyme, AOB directly use NH3 in the soil to produce NO2. By comparing the abundances and community shifts in the gene encoding for ammonia monooxygenase (amoA), we examined fundamental differences in these two seemingly similar global change factors on AOB, and therefore, nitrification.

We expected, based on results from the N-deposition treatment and the analogous nutrient pulse after burning, that AOB would respond to burning. In the N-deposition treatment we found an overwhelming increase in AOB abundance and a shift in community structure toward Cluster 3a, increased nitrification potential (Fig. 2c, d; Supplementary Table 2), and increased Gram-negative biomarkers (AOB are Gram-negative microorganisms). Calculated soil NH3 concentrations also increased in N-deposition plots (Fig. 2a), presumably the result of higher rates of N-mineralization resulting from increased plant productivity in N-deposition plots (Fig. 4b). These results are consistent with other studies that have shown increases in AOB abundance (e.g. Di et al. 2009; Shen et al. 2008; Schauss et al. 2009; Jia and Conrad 2009) and increases in the relative abundance of AOB Cluster 3a (Wang et al. 2009; Shen et al. 2008; Yeager et al. 2005; Bruns et al. 1999; Mendum and Hirsch 2002; Webster et al. 2002; Mintie et al. 2003; Horz et al. 2004; Kowalchuk et al. 2000a, b; Avrahami and Conrad 2003) in high N-NH3 soils. In one notable case, NH3 concentrations increased as a result of wildfire in a mixed conifer forest, which resulted in a long-term shift in the AOB community toward Cluster 3a ammonia-oxidizers (Yeager et al. 2005).
https://static-content.springer.com/image/art%3A10.1007%2Fs10533-011-9654-3/MediaObjects/10533_2011_9654_Fig4_HTML.gif
Fig. 4

Conceptual diagram describing the aboveground and belowground effects of N-deposition (left) and summer fire (right) over the growing season after treatment or fire. a The ambient control in April 2003, where plant productivity is limited by soil N and P. b depicts the imposed treatments. Left-hand side: Ca(NO3)2 was added to mimic dry deposition of nitrates, which led to an increase in nitrogen in the grassland system, and an alleviation of N-limitation, but not P-limitation. Right-hand side: we hypothesize that the summer fire treatment increased many soil nutrients, alleviating both N- and P-limitation of the plant community to a certain extent. c The effect of N-deposition on the ecosystem in the following growing season in April 2004. In general, plant biomass increased, which in turn increased litter and N-mineralization to NH3. This increased input of NH3 to the soil then promotes AOB activity, and allows for a shift in AOB community structure toward the high NH3-tolerant Cluster 3a sub-group. d The effect of fire on the ecosystem in the following growing season in April 2004. Again, plant biomass increases, presumably as a result of increased nutrient availability from the fire. However, the shift in AOB community and nitrification rates are not seen, despite higher soil NH3. This may be due to out-competition of AOB by the plants themselves, or by plants supplemented with additional capabilities by AMF

Given these results, we hypothesized that the similar increase in calculated soil NH3 concentrations exhibited in April 2004 in the burn plots (Fig. 2a) would affect AOB abundance and Cluster 3a relative abundance in similar stimulatory ways. Surprisingly, our results show that AOB abundance and community structure do not change at all in the burned plots (Fig. 2c, d). Furthermore, there was no difference in nitrification potential or in Gram-negative biomarker relative abundance between the unburned and burned plots in April 2004. This provides further indication that the AOB community was not stimulated by the increased calculated concentrations of soil NH3 resulting from the fire, or that stimulation occurred at a shorter or longer time scale than the first growing season post-fire. However, if AOB responses did exist at a shorter time scale post-fire, then our results suggest that AOB abundance and community structure recovered quickly within the 9-month period post-fire.

Plant competition for nutrients causes different microbial responses to wildfire induced nutrient pulses versus direct N-addition

Given the lack of response in AOB abundance and community to excess soil NH3, it is reasonable to suggest that the excess NH3 (calculated) present in the fire plots was not used for autotrophic nitrification as it was in the N-deposition plots (Fig. 2a). We propose instead that NH4+ in the fire plots is converted into plant AGB both directly via plant uptake, and indirectly through a possible mechanism involving AMF (Fig. 4). Although we report a trend of decreased AMF under nitrogen addition, which has also been reported previously for this system (Gutknecht et al. 2011), we observed no change in AMF after burning, despite a similar increase in calculated soil NH3. Since AMF have been shown to be stimulated by grassland fire (Hartnett et al. 2004), it is possible that AMF responses to fire are more immediate than our measurement 9 months post-fire could capture. Through growth of N-rich hyphae that increase plant root surface area, AMF can take up and transfer significant amounts of nitrogen and phosphorus to their host plants in both inorganic and organic forms (Abbott et al. 1984; Sanders and Tinker 1971; He et al. 2003; Govindarajulu et al. 2005; Hodge and Fitter 2010). There is evidence that plant phosphorus uptake increased after fire (Henry et al. 2006) and AMF associations could have provided their plant symbionts with a competitive advantage for that uptake after fire. Gram-positive bacteria, which were lower in the fire plots in April 2004 (Fig. 1), could be poorer competitors against AMF mainly for two reasons. First, because Gram-positive bacteria have higher N requirements for glycine peptide interbridges and higher P requirements for teichoic acid in their peptidoglycan layers, they have corresponding slow growth and thus cannot respond as quickly to nutrient pulses (Prescott et al. 1996). Also, bacteria in general, likely including AOB and other Gram-positive bacteria, would have been more sensitive than fungi to the decreased soil moisture after fire at our sites, as reported by Niboyet et al. (2011). While pH increased in the second (21 months) and third (33 months) years after fire, there was no response in the first growing season, so pH did not likely play a role in microbial responses for the first growing season after fire (Table 1; Niboyet et al. 2011).

It is also possible that shifts in AMF biodiversity, instead of total abundance, could be involved in increased plant primary productivity through enhanced nutrient uptake (Maherali and Klironomos 2007). Further analysis of the AMF community that is more specific than PLFA would be necessary to determine if this mechanism is active. This conceptual model potentially also explains the fire by N-deposition interactions seen in April 2004 and April 2005 in the soil bacterial community, where we saw higher relative abundance of Gram-positive bacteria (2004) or Gram-negative bacteria (2005) in N-deposition plots but not in N-deposition plots that were also burned (Fig. 4, Supplementary Fig. 1). This interactive effect has been reported at our sites for other measures of microbial nitrogen cycling (Niboyet et al. 2011), strengthening the idea that after burning, plants and plant-associated AMF were capable of out-competing soil bacteria for an influx of nutrients after burning. Unfortunately this conceptual model could not be verified with direct measures of phosphorus and nitrogen uptake by each component of our plant-microbial system, but this conceptual model offers potential mechanisms that could be tested in the future.

Conclusions

We examined the long-term effects of a low-severity fire in a California annual grassland ecosystem. In the first growing season following the fire, aboveground productivity increased as a result of fire, and litter decreased. However, we did not observe changes in belowground productivity or broad-scale microbial biomass. Instead, our results indicate that responses in specific portions of the soil microbial community occur as a result of fire, but that these responses do not persist past 3 years following fire occurrence. By examining even more specific microbial community members and functional genes using DNA/RNA sequencing, or metagenomics analysis, future studies may resolve microbial responses to fire still further. Also, though it was beyond the scope of this study, the decrease in Gram-negative lipids in April 2005 in the fire plots may be linked to a decrease in AOB abundance and nitrification rate, which could indicate longer-term effects of fire on soil microbial function and biogeochemistry.

Finally, we compared the fertilizing effect of the grassland fire to an N-deposition treatment in the same system, and found that autotrophic AOB are stimulated under N-deposition, but are not stimulated in fire treatments, despite similar impacts of the two treatments. We hypothesize that increased competition from plants and plant-associated AMF for N and P in the fire plots explains this difference in our observed effects on AOB.

In summary, although fire may act as a short-term disturbance to Mediterranean grassland ecosystems, the nutrient pulses specific to fire may affect microbial communities and above–belowground interactions in distinct ways. Our results indicate that microbial communities are more resilient to fire than aboveground plant biomass in this type of Mediterranean grassland. Overall though, we saw no distinguishable responses either above or belowground 33 months after fire, suggesting that these communities are resilient to this type of low-intensity grassland fire. As fire severity increases, due to increased deposition of atmospheric nitrogen or increased precipitation, it is unclear whether that resilience will continue in these systems.

Acknowledgments

The authors would like to thank our many collaborators at the Jasper Ridge Global Change Experiment. Specifically we thank Drs. Chris Field and Nona Chiarello for their work to keep the JRGCE continuing, for allowing our use of the above-ground, below-ground, and litter biomass data, and for their general support of our efforts as collaborators. Dr. Hugh Henry provided feedback on our interpretations of AGB data. We thank Yuka Estrada and Todd Tobeck for coordinating field sampling and sample processing. The Bohannan laboratory and specifically Dr. Sharon Avrahami provided methods and support of this project, as did the Balser laboratory, specifically Dr. Harry Read. We also thank reviewers of the manuscript for many helpful suggestions. Funding for this project was provided by the National Science Foundation (DEB 045-2652) and the NSF Postdoctoral Fellowship Program (Award Number 0805723).

Supplementary material

10533_2011_9654_MOESM1_ESM.pdf (24 kb)
Supplementary Fig. 1The interactive effect of fire and nitrogen deposition on the relative abundance (mol%) of a Gram-positive bacterial (15:0 iso) and b Gram-negative bacterial (16:1 ω7c) lipid indicators. Least squared means from ANOVA tests are presented for burned (white bars) and unburned (black bars) plots, and under ambient (left) and elevated (right) N-deposition treatments. Error bars represent one standard error associated with the statistical least squared means model. The interaction between fire and N-deposition was significant in April 2004 for Gram positive bacteria (F1,42.1 = 7.32, p = 0.010) and in April 2005 for Gram negative bacteria (F1,41.3 = 4.23, p = 0.05). (PDF 25 kb)
10533_2011_9654_MOESM2_ESM.pdf (32 kb)
Supplementary Table 1Summary of F-values, degrees of freedom, and p values from split-plot ANOVAs to test the treatment effects of fire and global changes on the following: aboveground biomass = log g m−2; belowground biomass = log g m−2; litter = log g m−2; pH = pH in water; percent soil moisture (%), and NH3 (2004 only) = calculated ammonia from extractable N-NH4+ (log M). 2003 values were based on all 128 plots of the JRGCE, while 2004 and 2005 data were based on the burn subset of data (see “Methods”). (PDF 33 kb)
10533_2011_9654_MOESM3_ESM.pdf (48 kb)
Supplementary Table 2Summary of F-values, degrees of freedom, and p values from split plot ANOVAs to test the treatment effects of fire and global changes on microbial indicators. 2003 values were based on all 128 plots of the JRGCE, while 2004 and 2005 data were based on the burn subset of data (see “Methods”). Microbial indicators: total microbial biomass (nmol lipid g soil−1); fungal:bacterial lipids = ratio of fungal to bacterial lipids; general fungi = 18:2 ω6,9c mol%; Gram-positive bacteria = 15:0 iso mol%; Gram-negative bacteria = 16:1 ω7c mol%; arbuscular mycorrhizal fungi = 16:1 ω5c mol%; AOB abundance (2004 only) = abundance of ammonia-oxidizing bacteria (copies of Bacterial amoA g dry soil−1); AOB Cluster 3a (2004 only) = community structure of ammonia-oxidizing bacteria (proportion of T-RF 434: total peak height); nitrification potential (2004 only) (ng N h−1 g dry soil−1). The ‘Multivariate’ column contains p values from Permanova nonparametric testing to determine treatment effects multivariately. Permanova analysis was performed on lipid relative abundance (mol%) data. The statistical model was based on the split-plot full factorial design of the JRGCE, run with 1000 permutations. (PDF 48 kb)

Copyright information

© Springer Science+Business Media B.V. 2011