Biogeochemical impacts of the northward expansion of kudzu under climate change: the importance of ecological context

Authors


  • Corresponding Editor: N. Barger.

Abstract

Climate change is generally expected to push plant species to higher latitudes and elevations; how the climate-induced migrations of disruptive invasive species may affect higher-latitude ecosystems has not been widely examined. Kudzu (Pueraria montana) has large impacts on nitrogen (N) cycling and trace N gas emissions in the southeastern United States. To understand how its projected northward migration under climate change will impact ecosystems in the northeastern United States, we examine the impacts of kudzu in the Mid-Atlantic region, near kudzu's northern invasion front. We pair plots invaded by kudzu with adjacent uninvaded plots, and examine rates of leaf litter decomposition, soil nitrogen pools and net cycling rates, N trace gas emissions, and microbial dynamics. Kudzu litter has more N and decomposes faster than litter from co-occurring species. Unlike in Georgia, near the center of kudzu's current range in the United States, kudzu invasion in the Mid-Atlantic has very small ecosystem impacts, causing significant increases only in the sizes of soil nitrate pools. These Mid-Atlantic ecosystems may be buffered against invasion impacts, creating a lag between changes in the plant community and biogeochemical changes. A combination of factors, including time since establishment, soil types, growing season length, and temperatures, may limit kudzu's biogeochemical impacts along its invasion front.

Introduction

Plant invasions and climate change are two major global change factors threatening ecosystems, and there may be important interactions between the two. Populations are generally expected to migrate to higher latitudes or higher elevations as warmer winters allow species to expand their ranges, and higher summer temperatures make lower latitudes or elevations less habitable (e.g., Kingsolver et al. 1992, Parmesan and Yohe 2003). One question in climate change's reshuffling of communities is how species—including invasive species—that play important roles in biogeochemical processes will respond to the changing climate (Walther 2009), and how this interaction between climate change and species invasions will affect ecosystem processes. While field, greenhouse, and modeling studies have examined the physiological and distributional responses of species and communities to changes in climate and atmospheric CO2 concentrations (e.g., Sasek and Strain 1988, Higgins and Harte 2006), the question of how species—particularly those with physiological characteristics likely to disrupt biogeochemical processes—will affect ecosystems as they expand into higher latitudes has not received the same attention.

Community composition and individual species are important drivers of ecosystem processes, along with other factors, such as soil and climate (e.g., Vitousek and Walker 1989, Hobbie 1992, Hooper and Vitousek 1998, Erickson et al. 2001). But community composition and species effects on ecosystem processes can vary depending on the environmental context. For example, outside of climate's direct effect on soil nutrient cycling (e.g., the direct effects of temperature on enzyme activity), it can also have indirect effects through its influence on broad and species-specific plant physiology, modulating community and species-specific impacts on ecosystems (e.g., temperature and freezing event effects on nodulation rates in legumes). In addition, ecosystems may be buffered against species-induced changes to ecosystem processes such as nutrient cycling, with lags in ecosystem responses to changes in community composition (e.g., Wardle et al. 1999, Compton et al. 2004, Latty et al. 2004, Hamman and Hawkes 2012). These interactions between climate and plant physiology and the buffering capacity of ecosystems are not as well studied as the independent direct effects of climate or species, but may be equally important. These effects can be particularly important in the application of ecosystem models to understand how plant invasions and changes in community composition are likely to affect ecosystems (e.g., Levine et al. 2006). If models do not incorporate buffering effects or the interactions between climate and the physiology of key invasive species that move northward with climate change (e.g., ecosystem buffering against N inputs or the effects of temperature on root nodulation rates), they may severely under- or over-estimate species and community effects on ecosystem processes.

Kudzu (Pueraria montana), a fast-growing leguminous vine native to Asia, is expected to move northward by hundreds of kilometers as global temperatures increase, and it is likely to have a range of ecosystem impacts throughout its new distribution (Sasek and Strain 1991). Vines are notoriously susceptible to freezing embolisms, and warmer winters are expected to allow kudzu to expand northward in coming decades (Wechsler 1977, Sasek and Strain 1991, Schnitzer and Bongers 2002). This expansion may be aided by increasing CO2 concentrations: vines in general have shown larger and more sustained growth responses to increased CO2 than trees (Hattenschwiler and Korner 2003), and the growth responses of N-fixers such as soybean have shown little acclimation to elevated CO2 (Ainsworth et al. 2002).

Currently estimated to be spreading by 50,000 ha yr−1, kudzu already covers over 3 million ha in the southeastern United States, roughly equivalent to the acreage of soybean agriculture in the region, making it the dominant nitrogen-fixer in the southeastern United States of America (USDA 2002, Forseth and Innis 2004). Kudzu exhibits a high degree of nodulation and nitrogenase activity in the United States, and a capacity for high rates of nitrogen (N) fixation has been observed in its native range (Lynd and Ansman 1990, Forseth and Innis 2004). Although fixation rates have not been measured in the U.S., kudzu is accelerating N mineralization and nitrification rates in Georgia soils, sometimes by an order of magnitude (Hickman et al. 2010).

Kudzu's impact may extend to the atmosphere by contributing to increased concentrations of tropospheric ozone, an important air pollutant in terms of its impacts on human health and agriculture (Hickman et al. 2010). Kudzu's physiology is unusual in that it can increase ecosystem-scale emissions of the key precursors to tropospheric ozone formation: it is a moderate- to high-emitter of the biogenic VOC isoprene, (C5H8) and has been shown to double soil emissions of nitric oxide (NO) (Sharkey and Loreto 1993, Hickman et al. 2010).

Though kudzu's expansion northward is well documented and understood (Wechsler 1977, Sasek and Strain 1991, Ainsworth et al. 2002, Lamont and Young 2004), and it has large impacts on N cycling and NO emissions in Georgia (Hickman et al. 2010), it is difficult to predict how kudzu will affect ecosystems in the northeastern United States. As described above, when kudzu expands into new areas, differences in climate and growing season length will alter its physiology and may moderate its inputs of N to ecosystems. Soils may be buffered against kudzu N inputs, and interactions with native litter may retard nutrient release during decomposition, so that biogeochemical impacts may lag considerably behind changes in community composition (e.g., Wardle et al. 1999, Compton et al. 2004, Hamman and Hawkes 2012, Hickman et al. 2013).

We examine how kudzu invasions affect a range of ecosystem properties at the northern edge of its distribution as a guide for understanding possible impacts from its future expansion into the Northeast. We expect the cooler temperatures that predominate in the northern part of kudzu's distribution to limit kudzu productivity and reduce N-fixation rates on a per unit area basis. These effects, in combination with the later establishment of kudzu in the Mid-Atlantic, are likely to reduce potential differences in N cycling and trace N gas production between invaded and uninvaded soils relative to the differences observed in the southern United States. While a decadal scale common garden experiment is essential for distinguishing climatic from buffering effects on kudzu's impacts, a field study such as is reported here is necessary to understand the scope and magnitude of these two important questions.

Materials and Methods

Sites

We selected sites in Maryland where winter temperatures tend to be colder than in the center of kudzu's current distribution and that are within the range that northeastern winter temperatures are expected to reach within the next 100 years. In Baltimore, MD, at a latitude where kudzu has been successfully established for decades, mean minimum January temperatures were 4.7°C from 1961 to 1990, roughly intermediate between means for southern sites where kudzu is established (−0.1° and 1.8°C in Athens, GA and Montgomery, AL), and means for northern sites within the range of kudzu's expected expansion (−9.8°C to −9.2°C in Poughkeepsie, NY, Worcester, MA, and Hartford, CT [NOAA/ESRL 2008]). Minimum temperatures in the northeastern United States are expected to increase by roughly 2–6°C by the end of the century, making northeastern winter temperatures similar to those currently experienced in Maryland (NCDC 2008).

We selected three sites where kudzu was well-established: McKee-Beshers Wildlife Management Area in western Montgomery County, MD (N 39 05.107, W 077 25.871), the Summit Hall Turf Farm, also in Montgomery County (N 39 05.315, W 077 26.556), and the Smithsonian Environmental Research Center (SERC) in Anne Arundel County, MD (N 38 51.945, W 076 33.815). Kudzu was well-established at SERC by 1975 (D. Whigham, personal communication), is believed to have established at McKee-Beshers before 1983 (K. D'Loughy, personal communication), and is believed to have established at Summit Hall before 1970 (D. Wilmot, personal communication).

Experimental design

The paired-site design has become common in studies of the effects of N-fixers on soil N studies and gas fluxes (e.g., Ashton et al. 2005). At each site, an invaded location was established within a stand of kudzu, and an uninvaded location was established in an area where kudzu had not invaded, within 40–200m of the kudzu stand, for a total of six paired locations. At SERC, the invaded location was within a former agricultural field where kudzu has been present for over 30 years. The uninvaded location was established on a former agricultural field abandoned 30 years ago, cleared 10 years later, and allowed to begin secondary succession during the last 20 years (D. Whigham, personal communication). Soils in both the invaded and uninvaded plots at SERC were a well-drained Marr-Dodon complex, derived from loamy fluviomarine deposits, with a slope of 2–5° (USDA 2008). Soils at all plots in McKee-Beshers and Summit Hall were well-drained Penn silt loam. At McKee-Beshers, both plots had 3–8° south-facing slopes; the plots at Summit Hall had 15 to 25 degree slopes (USDA 2008). The area adjacent to these plots has been managed as a flooded bottomland swamp for over 40 years; no management is conducted in the area where the plots are located (K. D'Loughy, personal communication). Plant cover at the sites was characterized in September 2005 and was described by Hickman and Lerdau (2006). In invaded locations at all three sites, kudzu cover ranged from 80% to 100%. Uninvaded locations at all three sites contained primarily herbaceous species, though the woody shrub Rosa multiflora was present in the uninvaded location at SERC. Although the defined uninvaded location contained only herbaceous species at McKee-Beshers, the surrounding area contained more and larger trees than were present in the uninvaded areas at SERC and Summit Hall. Senesced litter collected in litter traps (see below) was used to evaluate community differences in invaded and uninvaded sites. When kudzu is excluded from the genera of litter collected in the sites, there was no difference in the plant communities in invaded and uninvaded plots either on the basis of the presence/absence of genera (P = 0.277) or litter abundance by genera (P = 0.062).

Sample collection and preparation

Soil sampling was conducted nine times starting in March 2006: bimonthly from March through September in 2006 and 2007, and once in December 2006. At each sampling time, soil cores were taken from all sites within a single day, with the exception of September 2007, when soils from SERC were taken on 5 September 2007, and soils from the other sites were taken on 6 September. Three soil cores were taken randomly from an area measuring 5 m × 10 m in each location, for a total of 18 cores per sampling time (3 cores is a commonly-adopted sampling size in studies such as this [Martin et al. 2003, Hawkes et al. 2005, Haubensak and Parker 2004]). The one exception was sampling in September 2007, when four cores were taken per location, for a total of 24 soil cores. At sampling, the top litter layer was removed, and a PVC pipe (5 cm internal diameter × 20 cm long) was driven 12 cm into the ground and removed with the core intact. The cores contained little or no organic layer, so they were not separated into different horizons before being placed in separate polyethylene bags. The cores were kept cool until transferred to a refrigerator in the lab. Most lab analyses were started within three days of soil collection, but the denitrification assays were usually conducted four to six days after sampling, and analysis of soils from September 2007 were conducted approximately two weeks after sampling. Since precipitation can stimulate microbial activity (Paul and Clark 1996), samples were taken at least 24 hours after any major rain event.

Soils were homogenized by hand in Ziploc bags, and major rocks, roots, and invertebrates were removed. Subsamples from each core were taken for laboratory analysis of moisture content, total C and N, initial nitrate and ammonium, microbial biomass, net mineralization and nitrification potential, and denitrification enzyme assays (details below). Soil moisture content (g H2O g−1 dry soil) was determined by drying a subsample at 105°C until soils were no longer decreasing in weight. Total carbon (C) and N content of dried, ground samples was determined using a CE Flash EA 1112 Elemental Analyzer (CE Instruments, Milan, Italy).

Inorganic N pools

Inorganic N was extracted from soils by placing a 10 g subsample of soil from each core in a 120 ml polypropylene specimen cup with 50 ml 2M KCl, and shaking the cups for 60 minutes. The soils were allowed to settle for another 60 minutes after shaking. The KCl solution from each cup was filtered using Whatman #42 filter paper and frozen in a 40 ml glass scintillation vial until determinations of NO3/NO2-N and NH4+-N content were made using a Lachat autoanalyzer (Lachat Quickchem Systems, Milwaukee, Wisconsin, USA). These measurements also serve as the initial or “pre-incubation” inorganic N concentrations for the calculations of net N mineralization and net nitrification rates.

Net N mineralization and net nitrification

Concurrent with subsampling for the initial KCl extractions, a second 10 g soil subsample was taken and sealed in a mason jar with a gas-tight lid fitted with a rubber septum. The jars were incubated at 20–22°C. After 10 days, inorganic N was measured as described above; these extractions represent the “post-incubation” inorganic N content. Net mineralization was calculated as the difference in total inorganic N concentrations in the pre-incubation and post-incubation extractions; net nitrification was calculated similarly, as the difference in NO3-N + NO2-N concentrations in pre- and post-incubation soil extractions.

Soil microbial biomass

The determination of total microbial biomass was made using the chloroform fumigation incubation method (Voroney and Paul 1984). A 10 g subsample of soil from each core was fumigated for 12–18 hours. The 10 g samples of fumigated soils and a 0.1 g fresh soil inoculum from the same core were placed in a quart-sized mason jar and sealed with a gas-tight lid fitted with a rubber septum. After a 10 day incubation at 20–22°C, a 9 ml gas sample from the headspace of each mason jar was transferred to an evacuated glass vial. The vials were stored at room temperature until CO2 concentrations were determined by gas chromatography using a Shimadzu GC-14 GC fitted fitted with a thermal conductivity detector. Microbial biomass-C was calculated as the CO2-C per unit dry weight of soil in fumigated samples, divided by a constant (0.41) representing the fraction of biomass mineralized to CO2.

Denitrification enzyme activity

The denitrification potential of soils was determined using the Dentrification Enzyme Activity method (Smith et al. 1978). A medium containing 0.72 g KNO3 and 0.5 g glucose per liter of Nanopure H2O was created; 0.125 g choloramphenicol was added to inhibit microbial growth. A 5 g subsample was taken from each homogenized soil core and placed with 10 ml medium into a 125 ml Erlenmeyer flask with a ground glass neck and sealed with a rubber stopper. Flasks were made anaerobic with repeated evacuations and flushes with N2 gas, and 4 ml acetylene was added to each flask to inhibit the transformation of N2O to N2 by denitrifying bacteria. The flasks were placed on an orbital shaker, and 9 ml samples of the headspace of each flask were taken using a polypropylene syringe and transferred to evacuated gas vials after one and three hours. The N2O concentration of each vial was determined by gas chromatography using a Shimadzu GC-14 GC fitted with an electron capture detector. The change in N2O concentration from time zero through three hours was used to calculate the denitrification rates.

Trace N gas emissions

The invaded and uninvaded locations at SERC and Summit Hall were sampled for NO and N2O emissions on 5 and 6 September 2007. Gas measurements for each site were conducted on the same day, and all sampling was conducted between 10 am and 6 pm to limit variation in temperature between chambers. Sampling protocols were similar to those developed and described in Hickman et al. (2010). Briefly, in each invaded and uninvaded location, four beveled, Teflon-coated PVC rings (25.5 cm diameter) were randomly inserted several centimeters into the soil. At least 30 minutes after inserting the ring, a Teflon-coated, molded PVC chamber top fitted with a gas-sampling port was inserted over the ring and made gas-tight. Emissions of NO were measured in situ using a portable chemiluminescent analyzer equipped with a CrO3 filter that converts all NO to NO2 (Unisearch Associates model LMA-3D, Concord, Ontario, Canada). Standard curves were conducted in the field before and after each set of four measurements using a standard gas with a known NO2 concentration (0.01 ppm, Scott-Marin, Riverside, CA). Ambient NO2 concentrations were low but detectable, so NO2 concentrations within the chamber were measured immediately before and after NO measurements in order to measure the consumption of ambient NO2 by soils, which was assumed to be linear. NO emissions were measured as the linear increase in NO concentrations in the chamber over four minutes, and were corrected for the consumption of ambient NO2 during that four-minute period (Hall and Matson 2003).

To measure N2O emissions, a Teflon-coated, molded PVC chamber top fitted with a septum was placed over each ring and made gas-tight. Using polypropylene syringes, 9 ml gas samples were taken from the chamber at zero, 10, 20, and 30 minutes, and transferred to evacuated glass vials. The vials were stored at room temperature until analysis for N2O by gas chromatography using a Shimadzu GC-14 GC fitted with an electron capture detector. The N2O flux was calculated using the linear increase in N2O concentration, the chamber volume, and the soil surface area.

Litter decomposition

In October 2006, newly senesced leaf litter was collected from kudzu and 6 co-occurring woody species (Acer rubrum, A. negundo, Carya glabra, Fagus grandifolia, Quercus alba, and Sassafras albidum). Litter was collected along a transect approximately 500 m in length encompassing both invaded and uninvaded areas in the McKee-Beshers Wildlife Management Area, and running across Penn silt loam soils. Litter bags were constructed from nylon 1/32 inch (0.79 mm) square mesh, measuring 15.5 cm by 12 cm along the interior edges. Enough bags were constructed to allow for the destructive sampling of three replicates of each species at 18, 32, 44, and 53 weeks. Leaves were dried at room temperature, and 2.5 g of dried litter from a single species was placed in each mesh bag and weighed. Three 2.5 g air-dried samples from each species were weighed, dried at 60°C, and weighed again to calculate the weight difference between air-dried and oven-dried litter.

The litter bags were placed in the field 6 December 2006. Replicates were allocated randomly to each of three blocks located approximately 10 m apart from one another in an area of uninvaded forest in McKee-Beshers. Within a block, litter bags were placed randomly on a 3 m2 grid at 0.5 m intervals. At each harvest time, bags were collected, dried at 60°C, weighed, and ground for analysis of C and N content using a CE Flash EA 1112 Elemental Analyzer (CE Instruments, Milan, Italy).

Litter deposition

Five 30.5 × 61 cm litter traps were placed in each of the invaded and uninvaded locations in each site (30 litter traps total) in September 2007. Litter was collected from the traps weekly or bi-weekly through December 2007 and allowed to air-dry. Litter was separated by genus, dried at 60°C, and weighed. Subsamples of litter from each genus across multiple sampling dates were ground for analysis of C and N content using a CE Flash EA 1112 Elemental Analyzer (CE Instruments, Milan, Italy). The mean C and N concentrations for each genus was calculated for each site across sampling dates, and N inputs were calculated for each location by multiplying the mean N concentration for a genus by the weight of litter deposited by that genus during litter collection. A mean C:N for deposited litter in each location was also calculated.

Statistical analyses

Because of site disturbances in which plot markers were removed or destroyed between September 2006 and March 2007, it was necessary to establish new sampling locations within some sites. Consequently, two separate analyses were conducted: one for the 2006 growing season, and one for the 2007 growing season. Split-plot ANOVA's were conducted on mean response variables for each season, in which site was included as a whole plot factor, and kudzu invasion as a within plot factor. Single outliers were excluded from the analyses of net N mineralization and the NO3 pools in 2006, and of net nitrification in 2007.

Analyses of the NO and N2O fluxes were also conducted using split plot ANOVA, including site and kudzu invasion as whole and within plot factors, respectively. For the decomposition experiment, mass loss was analyzed in a three-way mixed model ANOVA, including block as the random factor, and species and harvest date as the fixed factors. The weight of N lost from litter over the first year was analyzed as a two-way mixed model ANOVA, including block as the random factor, and species as the fixed factor. When necessary, data were log transformed or rank transformed to meet the assumptions of ANOVA.

Results

The decomposition and nutrient dynamics of kudzu litter was clearly different from that of native species present at the sites in Maryland. Leaf litter from kudzu lost mass more quickly than six co-occurring tree species in 2007 (P < 0.0001, Fig. 1). By the end of one year, kudzu had lost most of its mass (56.2% ± 2.3% of starting mass), while on average, litter from each of the other species lost between 14.0% ± 1.0% and 38.0% ± 3.0% of its initial mass. Patterns of N loss, both as a percentage of starting N and in the net release of N to the soil, also showed significant differences between kudzu and the native species (P < 0.0001). While kudzu litter lost an average of 40.3% ± 2.38% of its starting N after one year, litter from all but one of the native species experienced net immobilization of N; on average, litter from native species accumulated 34.9% more N than at the start of decomposition, even as the mass of litter decreased (Fig. 1). In addition, kudzu had significantly higher starting concentrations of N, averaging more than twice as much N as the co-occurring tree species (2.56% ± 0.34% N for kudzu versus 1.08% ± 0.067% for native species).

Figure 1.

Mass loss (a) and N loss (b) during decomposition of leaf litter from kudzu and six co-occurring woody species in Maryland during 2007. Values are means ±1 SE.

Total input of N was higher in invaded plots than uninvaded plots (P = 0.05, one-tailed test, Table 1), but there were no differences in the mean litter C:N. The weight of kudzu litter input was higher in SERC than in the other two sites in 2007 (P = 0.004, Table 1).

Table 1. Comparison of litter chemistry and kudzu litterfall from September through December, 2007 in invaded and uninvaded sites; means are presented with standard errors in parentheses.Thumbnail image of

The differences in decomposition rates between kudzu and co-occurring species and the higher rates of N input in invaded plots have not yet contributed to consistent differences in soil N cycling and pools across the three Maryland sites during both years. Where differences were significant, the differences were in the direction expected for a scenario in which kudzu adds N to invaded soils: nitrate levels were 200% higher in invaded areas across all dates (P = 0.069 in 2006 and P = 0.074 in 2007; Fig. 2), net N mineralization and net nitrification rates were higher under kudzu in 2007, but not significantly so (P = 0.155 and P = 0.176 in 2007, respectively; Fig. 3), and denitrification enzyme activity was lower under kudzu in 2007, but also not quite significantly so (P = 0.11; Fig. 4). But kudzu invasion failed to affect net N mineralization, soil ammonium, total soil C, total soil N, microbial biomass, N2O fluxes, or NO fluxes across sites (Figs. 2, 3, 4, and 5). We were unable to test for interactions between site and kudzu presence, though visual inspection of the data suggests that there may have been some substantial effects of kudzu at SERC during 2007 for rates of net N mineralization, net nitrification, and nitrogen oxide emissions (Figs. 5 and 6). Soil moisture did not differ between invaded and uninvaded areas in either year.

Figure 2.

Soil pools of inorganic nitrogen at three sites in Maryland, from April, 2006 through September, 2007. Values are means ±1 SE.

Figure 3.

Net N mineralization (a) and net nitrification rates (b) at three sites in Maryland from April, 2006 through September, 2007. Values are means ±1 SE.

Figure 4.

Denitrification enzyme activity (a) and microbial biomass (b) at 3 sites in Maryland in 2006 and 2007. Values are means ±1 SE.

Figure 5.

Emissions of N2O (a) and NO (b) at two sites in Maryland in September, 2007. Values are means ±1 SE.

Figure 6.

Net N mineralization (left column) and net nitrification (right column) at each of the three sites in Maryland from April, 2006 through September, 2007. Values are means ±1 SE.

Discussion

With its much higher leaf litter N content and considerably faster rates of leaf litter decomposition than co-occurring species' (Fig. 1), kudzu is a picture-perfect example of a fast-cycling species invading a community with more recalcitrant litter and slower rates of nutrient cycling. Its litter has more than twice as much N as litter from co-occurring native species, and while litter from these native species exhibits substantial net immobilization of N over the course of a year, kudzu releases 40% of its initial N. Within the framework described by Hobbie (1992), a species such as kudzu that is able to cover large fractions of a landscape, and which has such strikingly different patterns of decomposition and nutrient release to the ecosystem, should be expected to reinforce faster rates of nutrient cycling as it becomes established in new ecosystems.

Although kudzu leaf litter is releasing N to soils at faster rates than the co-occurring species and invaded sites exhibited larger total N inputs in litter than uninvaded sites, kudzu invasion in Maryland has not yet resulted in the consistent increases in N cycling that have been observed in Georgia and that are consistent with the expected impacts associated with the invasion of a typical fast-cycling species. While net nitrification was higher in invaded soils in 2007, and NO3 pools were larger in invaded soils in both years, the effects were inconsistent. The different impacts in Maryland and Georgia highlight the importance of ecology, climate, and history in mediating the impacts that invasive species can have on ecosystem processes and the importance of considering context when developing models of invasive impacts. The inconsistent or delayed impacts of kudzu invasion on biogeochemistry contrast sharply with its impacts on community ecology. While community composition is directly and immediately altered by kudzu invasion (Hickman and Lerdau 2006), biogeochemical processes may change only after a significant time lag. The application of biogeochemical or other models using parameterizations based on community-ecosystem interactions at the heart of kudzu's range (or the center of any species' range) should be applied with caution in other sites, particularly where other factors may modulate the ability of species or communities to drive changes in ecosystem processes.

Rates of atmospheric N deposition in Maryland are slightly higher than in Georgia (Holland et al. 2005). While it is possible that higher rates of atmospheric N deposition common in the northeastern U.S. may accelerate N cycling in uninvaded soils and reduce or inhibit N fixation rates in N-fixers by increasing N availability in soils (Lucinski et al. 2002), this does not seem to be a likely explanation for the lack of a in impact of kudzu invasion in Maryland. We compared N cycling and pools in the uninvaded soils from our Maryland sites to those from uninvaded sites in Georgia. Net N mineralization, net nitrification, denitrification enzyme activity, and pool sizes of inorganic N did not differ in uninvaded soils in the two states (P = 0.56, P = 0.55, P = 0.28, and P = 0.51, respectively) for measurements taken six days apart in July and 10 days apart in September, 2007. However, the long history of N deposition in more northern states may make it a more important factor mediating kudzu's impact further north.

The ecological context within which an invasion occurs can strongly affect the consequences of the invasion. In the case of Maryland kudzu, the more nutrient-rich soils in Maryland as compared to Georgia may bias the system towards responding less strongly in terms of N-cycling impacts (Paul and Clark 1996, Potash and Phosphate Institute 2005). Similarly, the cooler climate (which can inhibit N-fixation activity) and shorter growing season of Maryland (which reduces the annual duration of N-fixation activity) can act in concert to minimize the magnitude and extent of N-fixation that occurs in the kudzu-invaded sites (Lindemann and Ham 1979, Fyson and Sprent 1982, Ryle et al. 1989, Peltzer et al. 2002, Robin et al. 2005, Houlton et al. 2008, Bedison and Johnson 2009).

Finally, and perhaps most importantly, simply by being closer to the invasion front, the Maryland kudzu populations are likely not to have been extant for as long, and any buffering capacity (e.g., Magill et al. 2004, Perakis et al. 2005) of the soils is that much less likely to have been exceeded. The combination of a cooler climate and shorter time since establishment may mean that kudzu stands in Maryland are not as dense as established stands in Georgia. Erickson et al. (2001) demonstrated that increased rates of N cycling in successional tropical forests were positively correlated to the abundance of naturalized N-fixing lianas, and a similar dynamic may explain some of the differences between Maryland and Georgia as well as among sites in Maryland. More generally, many belowground processes have been shown to lag considerably behind changes in plant community composition (Wardle et al. 1999, Habekost et al. 2008, Holtkamp et al. 2008, Hamman and Hawkes 2012). These lags between changes in community composition and subsequent changes in ecosystems processes are not yet commonly incorporated into invasion models, and they represent an important scientific frontier in the development of such models.

Although kudzu may not soon be the threat to northeastern ecosystems and regional air quality that it is in the southeastern U.S., it is impossible to dismiss the possibility that its establishment in the Northeast may be accompanied by future changes in N cycling and trace gas fluxes. Leaf litter from kudzu loses mass and N at higher rates than litter from co-occurring species, setting the stage for a transition from a slower- to a faster-cycling ecosystem. The possible differences in N transformations and trace gas emissions observed at SERC in 2007 and the sporadic increases in inorganic N pools in invaded soils at all sites may be indicative of ecosystems beginning this transition. Further studies that can distinguish the effects of time since invasion, stand density, and climatic and chemical effects on nitrogen fixation rates in kudzu will be important for better predicting its impacts as it migrates north.

Over time, the frequency of ecosystem impacts can be expected to increase as northeastern winters grow milder and growing seasons become longer, creating an environment more suitable to kudzu establishment and N-fixing activity. An increase of 2–6°C by the year 2100 would raise winter temperatures in Maryland above those currently experienced in Georgia, where kudzu's impact on N cycling is large and consistent. Elevated CO2 concentrations may enhance this effect, both by having a fertilizing effect on kudzu growth and by depleting soil N, giving kudzu a stronger competitive advantage over non-N-fixers. Though the full impact of kudzu on the ecosystems and atmosphere of northeastern states may not be experienced for years or decades after it invades, preventing its establishment will be essential to avoiding these potentially damaging impacts in the long term.

Acknowledgments

We thank Peter Groffman, Sharon Hall, and Blandy Experimental Farm for the use of facilities and equipment. This research was supported by the Garden Club of America and a National Science Foundation Dissertation Improvement Grant.

Ancillary