INTRODUCTION

Worldwide, grassland is being converted to woodland via either direct planting of woody plantations in grasslands or indirectly from alterations in climate or land use (e.g., grazing and fire regimes), which promote woody plant encroachment and proliferation in grasslands (Archer et al., 1995; Jackson et al., 2002; Briggs et al., 2005; Jackson et al., 2007). Regardless of the mechanism, such alterations in land cover impact many ecological processes. The effect this vegetation change has on ecosystem nutrient cycling and storage has been a major research focus for decades (e.g., Schlesinger et al., 1990). Interest in this topic remains high because conversion of grassland to woodland is occurring extensively worldwide, with implications for both current and future human land use (Archer et al., 2001). Furthermore, this vegetation change has the potential to sequester significant amounts of carbon (C) from the atmosphere into ecosystem pools (Pacala et al., 2001).

Recent studies indicate that aboveground net primary production (ANPP) increases with woody plant encroachment and proliferation in mesic grasslands (Harcombe et al., 1993; Reich et al., 2001; Hughes et al., 2006; Knapp et al., 2008). These large increases in plant production are likely to promote C accumulation over time, and increases in aboveground C pools (biomass and litter) are frequently reported at mesic sites undergoing woodland conversion (Hibbard et al., 2001; Jackson et al., 2002; McKinley and Blair, 2008). However, the response of belowground C pools, especially soil C, to this vegetation change has been harder to predict (Wessman et al., 2004). Woody plant encroachment and proliferation in grasslands can result in soil C accumulation (Boutton et al., 1998; Archer et al., 2001; Brandey and Young, 2010), no net change (McCarron et al., 2003; Smith and Johnson, 2003; Hughes et al., 2006), or loss (Jackson et al., 2002). Interestingly, mesic systems, where ANPP is thought to be most stimulated by grass-to-woodland conversion (Knapp et al, 2008), are also where soil C losses are more commonly observed with woody encroachment and plantation forestry (Guo and Gifford, 2002; Jackson et al, 2002).

How might woody plant dominance in mesic grasslands reduce soil C pools? There are several possible explanations: (1) Belowground net primary production (BNPP), a more direct source of C to soil pools than ANPP, may be reduced under woody plants (thereby reducing C inputs to soil C pools; Steinaker and Wilson, 2005); and/or (2) Woody plant dominance, through many possible mechanisms, may stimulate soil organic matter decomposition, for instance through changes in nitrogen (N) availability (Brandey and Young, 2008; Kirschbaum étal, 2008), alterations in microclimate (Vetaas, 1992), detritivore and microbial communities (Waldrop and Firestone, 2004), and/or a soil 'priming' effect (Fontaine étal, 2004).

The literature on woody plant encroachment and proliferation in grasslands contains many site-specific studies where one or more of these possible controlling factors on soil C gain or loss have been measured (e.g., Wessman et al., 2004). However, the majority of research has been done in xeric systems where either gains or no net change in soil nutrient pools have been observed (e.g., Hibbard et al., 2001; Lett et al., 2004; Smith and Johnson, 2004; Hughes et al., 2006; McClaran et al, 2008; McKinley and Blair, 2008). In an effort to identify some of die potential mechanisms governing soil C loss with woody plant proliferation in wetter grasslands, we measured plant inputs of C and N (ANPP and BNPP) and litter decomposition losses within a paired tallgrass prairie and adjacent woodland site in the soudiern Great Plains of North America shown previously to have experienced soil C losses with grass-to-woodland conversion (Jackson et al, 2002). We hypothesized diat while ANPP might be higher in woody than grass-dominated areas at this site, woodland BNPP would be significandy lower than native grassland. In addition, given dial some studies have shown grass-to-woodland conversion in mesic habitats increases N availability (Wilson and Kleb, 1996; Reich et al, 2001; McCulley et al., 2004; Brandey and Young, 2008), combined with die fact dial a potential N-fixing species [Prosopis glandulosa; (Johnson and Mayeux, 199O)] is dominant at this site, we hypodiesized that N inputs via plant production would be greater in the woodland dian die grassland. Finally, despite higher N inputs in die woodland, we predicted dial alterations in litter chemistry (e.g., more lignin/woody contributions) in the woodland would result in woody litter decomposing more slowly than grass material, as has been observed at other sites (Kochy and Wilson, 1997; Norris et al., 2001). These production and decomposition trends might mean dial over time C pools are shifted from primarily belowground in grassland to aboveground in woodland (into plant material and a developing litter layer; Jackson et al., 2002). Increasing N availability can stimulate decomposition of relatively labile material but may retard decomposition of recalcitrant litter or soil organic matter pools (Knorr et al., 2005). If soil C at this site is relatively labile in nature, the observed reductions in soil C pools that accompany woody plant establishment and dominance in this mesic grassland ecosystem might be the result of both reductions in belowground C inputs and increases in N input and availability, which may stimulate organic matter decomposition and loss.

MATERIALS & METHODS

SITE DESCRIPTION

We conducted this study in central Texas at the USDA-ARS field station near the city of Riesci, (31°29'N, 96°53'W), where annual precipitation is 896 mm (41 y mean on site) and mean annual temperature is 18.9 C. The annual precipitation at the site was 903 mm and 1171 mm for study years, 2003 and 2004, respectively. Using historical aerial photographs, supplemented with personal landowner accounts, we identified a location where native, uncultivated tallgrass prairie existed adjacent to (within 100 m) an uncultivated area that was originally grassland but that has become dominated by woody plant species (mesquite, Prosopis giandulosa, and others) over time, presumably due to changes in land management (e.g., cessation of mowing and burning). Carbon stable isotope signatures of the soil in the woodland confirm that the site was previously dominated by C4 vegetation (i.e., grasses: Jackson et a?, 2002). We established a 30 m X 30 m plot in each vegetation type based on similarities in soil type (Heiden clay; Udic Haplusterts) and topography [relatively flat (13% slopes) upland landscape positions].

The native grassland site has been managed as a hay pasture for many years and contains a plant community considered representative of native, remnant tallgrass prairie. Dominant grassland plant species included: Schizachyrium scoparium, Sorghastmm nutans, Desmanthus illinoensis, and Arnoglossum plantagineum (Policy et al., 2007). Long-term management of the site consists of annual Jun. mowing followed by hay removal, but the site receives no fertilizer applications. During the year of sampling, the grassland was hayed in Jun. of 2003 (and material was removed), immediately before this study began.

In contrast, the woodland has not been actively managed in >60 y and is dominated by seven woody species [Table 1; taxonomy according to (Diggs et al., 1999)]. Herbaceous vegetation in the woodland was relatively sparse and patchily distributed in the understory. An inventory of the woodland plot was performed in Sept. 2003 and consisted of recording the basal diameter of all live stems with dbh (diameter at breast height) >5 cm and counting the number of saplings >50 cm tall but with dbh <5 cm by species. Prosopis giandulosa was the dominant woody species at the site comprising >50% of the woody basal area (Table 1), although the two individuals with the largest basal diameters (33 and 34 cm) were both Gleditsia triacanthos. No Prosopis saplings were encountered in the woodland understory, suggesting this species may eventually be replaced by other co-occurring species. Using regression equations derived from Prosopis stands from a relatively nearby site (~400 km from Riesci) to relate basal diameter to age, the Prosopis stems measured in this study were 18-70 y old (Hughes et al., 2006). Using the oldest year in common of the three oldest trees [the method employed by Hughes et al. (2006)], stand age was estimated at 61 y old, which is in line with estimates from aerial photographs that indicate trees were present on site before 1941 (at least 62 y old at the time this study was conducted).

SOIL NUTRIENT POOLS

To verify changes in soil C and N pools with grass-to-woodland conversion at this site, we used a hydraulic soil probe to take ten 120 cm deep soil cores (4.0 cm i.d.) from random locations within each of the two 30 m X 30 m plots in the grassland and the woodland in Jun. of 2003. We used only non-compacted soil cores (n = 8 and 5, for grassland and woodland, respectively) for the following analyses. We sectioned the soil into five depth increments: 0-10, 10-30, 30-60, 60-90, 90-120 cm. We air-dried and sieved (2 mm) the soils, removing root fragments by hand. The soil was weighed and a sub-sample used for gravimetric water determinations. Bulk density was calculated as the water- and root-free mass of the soil divided by the volume. Additional sub-samples were ball-milled to powder, acid-treated to remove inorganic C from these marine sediment derived soils using a vapor method (Harris et ai, 2001), and analyzed for % C on a CE Instruments NC 2100 elemental analyzer (ThermoQuest Italia) . Percent N was determined in a similar manner, except on un-acidified samples, as acidification has been shown to alter measured N concentrations (Ryba and Burgess, 2002).

PLANT NUTRIENT INPUTS

We measured aboveground production (ANPP) using two techniques: peak standing crop in the grassland and litter tray collections in the woodland. We clipped five 50 X 50 cm quadrats in Sept. 2003, which gave us an estimate of aboveground grassland production since the Jun. 2003 haying event. We then clipped eight 50 X 50 cm quadrats in late Aug. 2004, and subtracted the Sept. 2003 average biomass from the individual Aug. 2004 quadrat biomass data for aboveground grassland production estimates (Sept. 2003-Aug. 2004). Although the peak standing crop technique may be less accurate than more frequent sampling methods that track changes in live and standing dead material throughout the year, it gives accurate results in humid temperate grassland systems, such as the site measured in this study (Scurlock et al, 2002). We coupled this once-a-year grassland sampling scheme with litter collections that occurred every 2-3 mo from 10 trays (34 cm X 50 cm each) placed randomly throughout the woodland site. Trays were installed JuI. 2003, and all material in them (e.g., seed pods, leaves, twigs, etc.) was collected in Sept. and Nov. 2003 and in Feb., May, Jun., and Sept. 2004. For the annual total woody litter production estimates, we summed the Nov. 2003 (which included Oct. and Nov. litter) through Sept. 2004 values. We also clipped four 50 X 50 cm quadrats in the woodland in late Aug. 2004 for estimates of understory herbaceous production. The average biomass and elemental content of this understory vegetation (on a g m"2 basis) was added to that of the woody litter production estimates to derive total woodland aboveground production values. We were unable to estimate wood production on the main trunk and limbs of the trees in this study; however, we checked our litter tray productivity estimates with allometry-derived aboveground productivity equations developed for Prosopis stands at a nearby site (Hughes et al., 2006) and found surprising similarity between the two (see Results for more details). Other studies have also shown that annual radial growth of woody stems accounts for a small proportion of aboveground tree litter production (<5%, Stienaker and Wilson, 2005).

Harvested litter and biomass was oven-dried for 5 d (50 C) and weighed. The entire sample was ground in a cyclone mill, and a sub-sample was then ball milled and analyzed for C and N content on the same elemental analyzer used for the soil analysis. We calculated the nitrogen use efficiency (NUE) of both vegetation types as the aboveground plant production (g m~2) divided by the amount of N in litterfall (a.k.a., the N contained in the Aug. harvest period for the grassland and the combined total of the litter collections for the woodland) .

We used a modified root in-growth technique to measure belowground production in both the grassland and woodland systems (Lauenroth, 2000) . In JuI. 2003, we excavated eight randomly placed 0-30 cm deep holes in both the grassland and woodland sites using a bucket augur (11 cm i.d.). We refilled these holes with root-free silty clay soil (Austin soil series containing 1.84% organic C and 0.16% N) collected from a nearby site. We collected root in-growth samples in Sept. 2003, Jun. 2004, and Sept. 2004, by driving a 4 cm i.d. soil corer into the center of the excavated hole and extracting one 0-30 cm depth increment. After each sample extraction, more root-free soil was used to fill the holes.

Extracted root in-growth samples were oven-dried (5 d at 50 C) for storage and transportation purposes. Within 6 mo of field collection, roots were removed using a flotation and hand-picking procedure, where DI water was added to sub-samples of each 0-30 cm core to soften the soil and was then sorted and picked until all root segments >2 mm in length were extracted from the total core. The roots were then dried, weighed, ground in a ball mill, and analyzed for C and N content. Samples that contained enough mass were further subdivided for ash determination (combusted at 500 C for 5 h in a muffle furnace). Belowground production values were calculated as the ash-corrected sum of the Jun. and Sept. 2004 sampling dates.

LITTER DECOMPOSITION

We measured decomposition rates of the dominant vegetation in both the grassland and woodland using litterbags filled with aboveground plant material collected in Sept. 2003. Aboveground material in the grassland was collected by clipping five 50 X 50 cm randomly placed quadrats. All clipped material was dried (5 d at 50 C), cut into 2-4 cm pieces with grass shears, and mixed well throughout litterbag construction. For the woodland, we removed ~1 m long healthy limbs from the dominant species. These limbs were sectioned into smaller pieces, dried (5 d at 50 C), and then separated into leaf and wood components (by species). Small (<2-cm diameter) twigs and stems were cut into discs and small pieces. Although we did not use senesced material for our decomposition experiment, litter bag C:N ratios (51 for grassland and 25 for woodland) were reflective of overall above- and belowground production C:N differences between the two litter types (averaged above- and belowground C:N values were 72 vs. 32 for grassland and woodland, respectively; Fig. 2) .

Litterbags (10 X 10 cm2) constructed of fiberglass-nylon mesh (1X2 mm2 openings) were packed with 3.79 g of material. Both grassland and woody litterbags contained a mixture of species present in each site. The grassland bags were constructed from the wellmixed composite clipped material. Woody litterbags contained a mixture of the dominant species leaf and wood material based on the proportion of total basal diameter covered by each species (Table 1; material from species with <3% basal area coverage was not included) and from a previous study in /Vraopis-dominated woodland that showed litterfall was 60% leaf material and 40% wood (McCulley, 1998). We assumed a 60/40 leaf-to-wood litter production ratio for all woody species.

A reciprocal transplant experimental design was used with the litterbags to determine if differences in microclimate or decomposer communities between the two vegetation types were important in controlling litter decomposition rates. Grass and woody litterbags (n = 66 of each) were randomly placed on the soil surface in Nov. 2003 in both the grassland and woodland locations and were harvested in Feb., May, Jun., and Sept. 2004 (4-6 litterbags of each grass and woody litter type were randomly collected per location and harvest date). Initial mass and elemental content was determined from litterbags (n = 8 and 9 for grass and woody bags, respectively) taken to the field in Nov. 2003 but brought immediately back to the lab. Initial and harvested litterbags were dried (5 d at 50 C) ; litter was removed from the bags and weighed. We ground the material using a cyclone and ball mill and determined C, N, and ash content using previously described methods. Because litterbag samples can accumulate significant quantities of soil over time, we also soil-corrected our C & N litter values (Hunt et al, 1999) . We calculated the decomposition constant [on a mass (k, y~ ') and carbon (kc, y"1) basis] over the study period by fitting the following equation to the data:

ln(X,/X")=-kt

where X, and X0 are the ash-free mass or carbon content of the litter at the final sampling period (Sept. 2004; i = 0.83 y"1) and time O (Nov. 2003) (Olson, 1963). In order to evaluate differences in N dynamics, we calculated the ratio of N remaining in the litter at each harvest period to that of the initial litter (on a mass basis; Nt/N0) .

STATISTICAL ANALYSES

We used several statistical approaches to assess the effect of grassland conversion to woodland on the measured parameters. Because depth increments within individual soil cores are correlated, we ran a multivariate analysis of variance (MANOVA) test using a general linear model (SAS, 1996) to determine vegetation type differences in the soil profile % C, % N, and bulk density data. For the total 120-cm-deep soil profile C and N content (g C or N m~2) and the above- and belowground production data, we utilized a simple paired i-test procedure employing the Satterthwaite variance calculation to test for differences between vegetation types. To assess litter and habitat type effects on litter decomposition rates (k and kc) and N dynamics accompanying decomposition, we ran a repeated-measures mixed model using harvest date, litter type (grass or woody material), and location (incubated in grassland or woodland) as fixed effects and the rep (litter type X location) effect as a repeated component. For all statistics, individual soil cores, quadrats, litter collection trays, and/or litter bags from each vegetation type were considered replicates. Ideally, multiple paired sites of woodland and adjacent grassland would have been identified and used as replicates in this study; however, in this geographic region (an area where intensive agriculture has historically and still is widely practiced) , finding adjacent sites of unplowed native grassland and >50 y old mesquite dominated woodland was not possible. Normality and equality of variance assumptions were checked prior to analyses of all dataseis, and appropriate transformations and/or selection of statistical methods were made accordingly.

RESULTS

SOIL NUTRIENT POOLS

Consistent with previous sampling at this site (Jackson et al, 2002), total soil C pools were reduced in woodland compared to adjacent grassland by ~20% (Fig. 1). This trend, while marginally insignificant (P = 0.06), reflects the fact that organic C concentrations tended to be reduced in woodland soils, and grassland soil bulk densities were significantly greater than adjacent woodland, perhaps resulting from the mechanized haying practices employed in the grassland (Fig. 2). There was no significant difference between vegetation types in total soil N pools (Fig. 1), but the concentration of N was higher in the woodland surface soil (0-10 cm; Fig. 2). O.N ratios of both the surface soil and the total 0-120 cm depth profile were significantly lower in woodland than adjacent grassland (Figs. 1 & 2).

PLANT PRODUCTION

Consistent with the traditional dogma that grasses allocate proportionally more C belowground than woody species, root production in woodland was less than half that of adjacent grassland (Table 2) and accounted for a smaller proportion of the total C inputs to the ecosystem (10.4% vs. 23.2% for woodland and grassland, respectively). However, %C of woodland roots (35.7%) tended to be higher than that of grassland (32.3%) (P = 0.0557). Total carbon inputs via plant production were surprisingly similar in grassland and woodland vegetation types at this site: net primary production was 330 and 342 g C m^sup -2^ y^sup -1^ for grassland and woodland, respectively. There was no significant difference in aboveground production between the vegetation types (Table 2; Fig. 3) , regardless of whether or not understory herbaceous aboveground production (64 g C m^sub -2^ y^sup -1^) was included in the woodland estimate. Because the litter tray technique we employed to measure woody aboveground production does not quantify radial wood production or production of new stems or branch material, we compared our data with estimates derived from regression equations relating stand age to aboveground productivity from a relatively nearby AiosojNs-dominated site (Hughes et al., 2006). Our aboveground woody production values were surprisingly similar to those derived from the allometric equations [244 vs. 206 g C m^sup -2^ y^sup -1^, this study vs. Hughes et al. (2006), respectively].

In contrast to the C input patterns measured between the two vegetation types, N inputs were significantly greater in the woodland than the grassland, consistent with the fact that mesquite is a N-fixing species. Nitrogen inputs measured in woodland total production were 3× that of adjacent grassland (Fig. 3), and significant differences between vegetation types were more apparent above- than belowground. Woodland litterfall peaked in the fall (Sept.Nov.), with woody litter N concentrations also lowest at this time of the year (~1.1%N) though still significantly higher than grassland fall biomass %N (0.6%; data not shown). Percent N in woodland roots (1.37%) was significantly higher than that of grassland (0.51%; P < 0.0001). As a consequence of these trends, C:N ratios of both above- and belowground production inputs in woodland were half that of grassland (Fig. 3), and the NUE of grassland ANPP (g biomass/g N in litterfall) was 2X that of woodland (P < 0.0001; Table 2).

LITTER DECOMPOSITION

Litterbags composed of grass and woody material had initial C:N ratios of 51 and 25, respectively, reflecting the observed differences in C:N of the plant production measurements from the different vegetation types. Yet, despite this large difference in initial litter chemistry, -40-50% mass remained in both types of litterbags after almost a year in the field (Fig. 4), and there was no significant 'litter type' main effect on calculated litter decomposition rates (k or R1) (Table 3), indicating that these two very different types of litter decomposed from a mass and C perspective in a surprisingly similar manner. This was not the case with regard to N. The 'litter type' main effect was highly significant in the repeated measures ANOVA on N retention (P = 0.0001), as was the 'month*litter type' interaction (P = 0.0002). Grass litter retained significantly more N than the woody litterbags (Fig. 4c), and the two litter types exhibited different patterns of N retention over time. Woody litter mineralized N throughout the experiment; whereas, grass material tended to retain it. These trends were also reflected in the C:N ratio of the remaining litter material (Fig. 4d). Grass litter C:N ratios declined throughout the experiment, while woody litter C:N ratios remained relatively constant.

The 'location' of litterbags in the field (i.e., whether litterbags were incubated in the grassland or woodland site) was not significant for k, k,., N retention, or C:N ratios; however, 'location' was significant in the mass and C loss repeated measures ANOVA, and the 'location*litter type' interaction was significant for both of these parameters and K0 and N retention. For mass, C loss and k,-, the significant interaction term was primarily the result of each litter type losing mass and C more quickly when incubated in its 'native' location (Table 3; Fig. 4a, b) . The grass litterbags incubated in the woodland location decomposed the slowest from both a mass and C perspective, especially early on in the decomposition study (Fig. 4a, b), and contributed to the significant 'location' effect in the repeated measures ANOVA. With N retention, both litter types retained more N when incubated in the 'non-native' location (Fig. 4c). These results suggest that litter decomposition as a process has differentiated between these two vegetation types, despite their close proximity and similar macroclimate and soil type, to favor the decomposition of the dominant vegetation type.

DISCUSSION

Reduced belowground plant production and enhanced N cycling rates, as evidenced by higher surface soil %N concentrations, higher N content and lower NUE of aboveground litter inputs, and less N retention in decomposing woody litter, may contribute to the loss of soil C observed to accompany woody plant proliferation and dominance in this mesic tallgrass prairie. Reductions in soil C pools, while marginally insignificant, were comparable with the scale of the reduction (-20% less soil C in woodland than grassland) previously reported from this site (Jackson et al., 2002), and represent a large amount of C (4.6 kg Cm^sup -2^). Soil C represents the balance of C inputs via plant production and C outputs via soil organic matter and litter decomposition and dissolved organic C loss integrated over time, all of which likely vary as woody stands age, and species composition and/or climate and land use changes. This temporal variability in the controlling parameters presents a major challenge to understanding grass-to-woodland conversion effects on soil C pools. While we examined only one site, our data, showing that woodland belowground plant production, at least in the top 30 cm of soil, is half that of the adjacent grassland and that aboveground litter C deposition and decomposition rates are similar between the two vegetation types, suggests that woodland soil C pools are unlikely to increase in the near future, and they may continue to decline. Similar to prior studies (e.g., Jackson et aL, 2002), these results suggest that ecosystem C storage is shifting aboveground in die conversion of grassland-to-woodland at this site. It is likely that woody species have deeper root production than was quantified in this study (30 cm). Additional work at the site with stable C isotopes has shown significant woody plant contributions to soil organic C pools to ~2 m in depth but not below (Jackson et al, 2002); however, both this prior work and our study indicate that despite these deeper woody plant root contributions, total soil organic C pools at depth are similar between grassland and woodland sites.

Reduced belowground production in woodland compared to grassland, as we measured here, is consistent with previous studies showing greater root biomass in grass dominated systems (Jackson et al., 1996). Though our BNPP values are considerably lower than those reported for prairie and adjacent aspen forests in Canada (Steinaker and Wilson, 2005), both our study and that of Steinaker and Wilson (2005) found no significant difference in the amount of N in BNPP between the vegetation types but measured 3× larger N inputs via ANPP in woodlands than adjacent grassland. Alterations in N cycling have been implicated as a potential mechanism enhancing plant invasions, such as trees moving into grassland (Laungani and Knops, 2009). Increased rates of N cycling with woody plant dominance in grasslands has been reported for several woody species: mesquite (Prosopis glandulosa, Schlesinger and Pilmanis, 1998; Hibbard et al., 2001), aspen (Pofwlus tremuloides, Wilson and Kleb, 1996), oaks (Queráis spp., Reich et ed., 2001), and wax myrtle (Morella cerifera, Brantley and Young, 2008). Some of these species, including the dominant tree in our study, mesquite (Prosopis), are capable of forming symbiotic N-fixing associations with various microbes (Johnson and Mayeux, 1990; Brantley and Young, 2008). In the case of oaks (and potentially junipers), reductions in fire frequency that typically co-occur when these woody species invade grassland may reduce N volatilization losses and lead to a gradual build up of N pools and eventual increases in N availability in developing woodlands (Reich et al., 2001; McKinley and Blair, 2008) . Although we did not assess N-fixation in the mesquite present at our site, elevated %N in surface soils, lower C:N ratios throughout the soil profile, higher %N in both roots and aboveground litter, lower aboveground NUE, and less N retained in decomposing litter, all suggest that N is more available and cycling more quickly in the woodland at this site. Enhanced N availability can have varying effects on litter and soil organic matter decomposition, depending on the chemistry of the material being degraded, overall nutrient availability at the site, and die type of microbes involved in the process (Janssens et al., 2010). It is probable that enhanced N availability directly or indirecdy alters ecosystem processes that ultimately control soil C pools (Kirschbaum et al., 2008). Additional process oriented studies that further explore these possibilities at this site seem warranted.

The fact that aboveground litter decomposition rates were similar, on a mass and C loss basis, between the two vegetation types was surprising; given the large and significant differences in initial C:N ratios of the litter. This result contradicts basic decomposition theory that higher C:N litter should decompose more slowly than lower C:N material (Aerts, 1997), and differs from previous work that has shown grass litter decomposes more quickly dian juniper and aspen litter (Kochy and Wilson, 1997; Norris et al, 2001). We did not measure lignin content of our litter, and given that wood was included in our woody litter bags, it is possible dial higher lignin content in woody litter bags interacted with lower C:N ratios to keep overall decomposition rates similar between the two litter types. The significant 'location' effect in the repeated measures ANOVA for mass and C loss concurs with prior studies on this subject that decomposition tends to occur more rapidly in grassdominated systems (Kochy and Wilson, 1997; Norris et al., 2001). However, it is notable that this 'location' effect was primarily driven by the slow rate of grass litter decomposition in woodland and was not observed in the statistical analysis of k or k,- parameters, perhaps suggesting that differences in decomposition rates between the two vegetation type locations were not that strong. We did not quantify differences in biotic (arthropod or microbial) communities or environmental conditions between the two vegetation types, but such differences may have contributed to the observed result that each litter type tended to decompose more rapidly when incubated in the 'native' habitat/location. Interactions between litter quality and habitat-specific biota and microclimate may promote similar litter decomposition rates in grassland and woodland at this site.

Perhaps the most surprising result of our study is that ANPP was similar in our grassland and woodland sites. This finding differs markedly from recent reports of dramatic increases in ANPP accompanying woody plant dominance in other mesic grasslands of North America (Knapp et al., 2008). According to the Knapp et al. (2008) regression of mean annual precipitation versus increases in ANPP with woody plant dominance in grassland, we should have seen ~800 g m~2 more production in the woodland than grassland at our site. Instead, we measured ~100 g m~2 difference between the two, which was not significantly different from zero. While our measured woodland ANPP value is within the range reported in Knapp et al. (2008) for other mesquite dominated sites, our grassland ANPP is significantly higher than the mesic grasslands included in that work. Yet, our grassland ANPP data agree with other independent estimates of ANPP at this Riesel, TX location (Polley et al., 2007) and are within the range of other humid temperate grasslands worldwide (Scurlock et al., 2002). Riesel occurs in the 'Blackland Prairie' of Texas, an area characterized by high clay, deep, fertile soils, warm temperatures, and relatively high rainfall (Smeins and Diamond, 1983) . The large amount of grassland ANPP at this site may be the product of both the abiotic conditions and the relatively low intensity haying management practices employed. It is possible that ANPP between the two vegetation types may have differed more in unusually dry or wet years. It seems likely that differences in abiotic and edaphic conditions, ages of the woody stands, species involved, and/or management practices of the various sites account for this discrepancy between our work and that of Knapp et al. (2008).

Effects of woody plant encroachment and proliferation in grassland on ecosystem processes can be significant and vary over time and on a site-by-site basis. This variability is probably related to differences in species characteristics (of both the woody invaders and the grassland species they are invading) and the abiotic and edaphic conditions and management history of the sites. More complicated, multifactor models, that take these parameters into account, may be required to accurately represent changes in ecosystem parameters that accompany woody plant encroachment and proliferation in a variety of grassland systems and to help develop a better understanding of when gains and losses of C will accompany this type of vegetation change.

Acknowledgments. - The USDA-ARS Grassland Soil and Water Research Laboratory in Temple, TX generously provided access to the grassland field site and lab space and equipment necessary for the research. We thank Mrs. Adeline Finski of Riesel, TX for allowing us access to the woodland site. We thank Anne Gibson, Kate Budd, Will Cook, Jim Nelson, Hyrum Johnson, and Wayne Polley for their input and assistance throughout this project. This work was supported with funds provided by the USDA-National Research Initiative (#2005-35101-17371) and the Kentucky Agricultural Experiment Station (KVD06045) to R.L.M. and Department of Energy's National Institute of Global Environmental Change and National Institute for Climate Change Research awards to R.B.J.

LITERATURE CITED

AEKTS, R. 1997. Climate, leaf litter chemistry and leaf litter decomposition in terrestrial ecosystems: a triangular relationship. Oikos, 79:439-449.

ARCHER, S., D. S. SCHIMEL AND E. A. HOLLAND. 1995. Mechanisms of shrabland expansion: Land use, climate or CO2? Ctimat. Change, 29:91-99.

_____, T. W. BOUTTON AND K. HIBBARD. 2001. Trees in grasslands: biogeochemical consequences of woody plant expansion, p. 115-138. In: E. D. Schulze, M. Heimann, S. Harrison, E. A. Holland, J. Lloyd, I. Prentice and D. S. Schimel (eds.). Global Biogeochemical Cycles in the Climate System. Academic Press, San Diego, California.

BOUTTON, T. W., S. ARCHER, A. J. MIDWOOD, S. F. ZITZER AND R. BOL. 1998. 513C values of soil organic carbon and their use in documenting vegetation change in a subtropical savanna ecosystem. Geoderma, 82:5-41.

BRANTLEY, S. T. AND D. R. YOUNG. 2008. Shifts in litterfall and dominant nitrogen sources after expansion of shrub thickets. Oecologia, 155:337-345.

_____ AND _____. 2010. Shrub expansion stimulates soil C and N storage along a coastal soil chronosequence. dob. Change Biol., 16:2052-2061.

BRIGGS, J. M., A. K. KNAPP, J. M. BLAIR, J. L. HEISLER, G. A. HOCH, M. S. LETT AND J. K. MCCARRON. 2005. An ecosystem in transition: Causes and consequences of the conversion of mesic grassland to shrubland. BioScimce, 55:243-254.

DIGGS, G., B. LIPSCOMB AND R. O'KENNON. 1999. Shinners and Mahler's Illustrated Flora of North Central Texas Botanical Research Institute of Texas. Fort Worth, Texas.

FONTAINE, S., G. BARDOUX, D. BENEST, B. VERDIER, A. MARIOTTI AND L. ABBADIE. 2004. Mechanisms of the priming effect in a savannah soil amended with cellulose. Soil Sd. Soc. Am. J., 68:125-131.

GUO, L. B. AND R. M. GIFFORD. 2002. Soil carbon stocks and land use change: a meta analysis. Glob. Change Biol, 8:345-360.

HARCOMBE, P. A., G. N. CAMERON AND E. G. GLUMAC. 1993. Above-ground net primary production in adjacent grassland and woodland on the coastal prairie of Texas, USA. /. Veg. Sci., 4:521-530.

HARRIS, D., W. R. HORWATH AND C. VAN KESSEL. 2001. Acid fumigation of soils to remove carbonates prior to total organic carbon or carbon-13 isotopie analysis. Soil Sa. Soc. Am. ]., 65:1853-1856.

HIBBARD, K., S. ARCHER, D. S. SCHIMEL AND D. W. VALENTINE. 2001. Biogeochemical changes accompanying woody plant encroachment in a subtropical savanna. Ecology, 82:1999-2011.

HUGHES, R. F., S. ARCHER, G. P. ASNER, C. WESSMAN, C. R. McMuRTRY, J. A. NELSON AND R. J. ANSLEY. 2006. Changes in aboveground primary production and carbon and nitrogen pools accompanying woody plant encroachment in a temperate savanna, dob. Change BwL, 12:1733-1747.

HUNT, H. W., D. E. REUSS AND E. T. ELUOTT. 1999. Correcting estimates of root chemical composition for soil contamination. Ecology, 80:702-707.

JACKSON, R. B., J. L. BANNER, E. G. JOBBAGY, W. T. POCKMAN AND D. H. WALL. 2002. Ecosystem carbon loss with woody plant invasion of grasslands. Nature, 418:623-626.

_____, K. A. PARLEY, W. A. HOFFMAN, E. G. JOBBAGY AND R. L. MCCULLEY. 2007. Ch. 19: Carbon and water tradeoffs in conversions to forests and shrublands, p. 237-246. In:]. Canadell, D. E. Pataki and F. Louis (eds.). Terrestrial Ecosystems in a Changing World. Springer, Cambridge, Massachusetts.

_____, J. CANADELL, J. R. EHLERINGER, H. A. MOONEY, O. E. SALA AND E. D. SCHULZE. 1996. A global analysis of root distributions for terrestrial biomes. Oecologia, 108:389-411.

JANSSENS, I. A., W. DIELEMAN, S. LUYSSAERT, J.-A. SUBKE, M. REICHSTEIN, R. CEULEMANS, P. CIAIS, A. J. DOLMAN, J. GRACE, G. MATTEUCCI, D. PAPALE, S. L. PIAO, E.-D. SCHULZE, J. TANG AND B. E. LAW. 2010. Reduction of forest soil respiration in response to nitrogen deposition. Nat. Geosci., 3:315-322.

JOHNSON, H. B. AND H. S. MAYEUX. 1990. Prosopis giandulosa and the nitrogen balance of rangelands: extent and occurrence of nodulation. Oecologia, 84:176-185.

KIRSCHBAUM, M. U. F., L. B. Guo AND R. M. GIFFORD. 2008. Why does rainfall affect the trend in soil carbon after converting pastures to forests?: A possible explanation based on nitrogen dynamics. For. Ecol Manag., 255:2990-3000.

KNAPP, A. K., J. M. BRIGGS, S. L. COUJNS, S. ARCHER, M. S. BRET-HARTE, B. E. EWERS, D. P. C. PETERS, D. R. YOUNG, G. R. SHAVER, E. PENDALL AND M. B. CLEARY. 2008. Shrub encroachment in North American grasslands: Shifts in growth form dominance rapidly alters control of ecosystem carbon inputs, dob. Change Biol., 14:615-623.

KNORR, M., S. D. FREY AND P. S. CURTIS. 2005. Nitrogen additions and litter decomposition: A metaanalysis. Ecology, 86:3252-3257.

KOCHY, M. AND S. D. WILSON. 1997. Litter decomposition and nitrogen dynamics in aspen forest and mixed-grass prairie. Ecology, 78:732-739.

LAUENROTH, W. K. 2000. Methods of estimating belowground net primary production, p. 58-71 . In: O. E. Sala, R. B.Jackson, H. A. Mooney and H. W. Howarth (eds.). Methods in Ecosystem Science. Springer-Verlag, New York, New York.

LAUNGANI, R. AND J. M. H. KNOPS. 2009. Species-driven changes in nitrogen cycling can provide a mechanism for plant invasions. P. Nati. Acad. Sci. USA, 106:12400-12405.

LETT, M. S., A. K. KNAPP, J. M. BRIGGS AND J. M. BIAIR. 2004. Influence of shrub encroachment on aboveground net primary productivity and carbon and nitrogen pools in a mesic grassland. Can. J. Bot., 82:1363-1370.

MCCARRON, J. K., A. K. KNAPP ANDJ. M. BLAIR. 2003. Soil C and N responses to woody plant expansion in a mesic grassland. Plant Soil, 257:183-192.

McCLARAN, M. P., J. MOORE-KUCERA, D. A. MARTENS, J. VAN HARÉN AND S. E. MARSH. 2008. Soil carbon and nitrogen in relation to shrub size and death in a semi-arid grassland. Geoderma, 145:60-68.

MCCULLEY, R, L. 1998. Soil respiration and microbial biomass in a savanna parkland landscape: spatiotemporal variation and environmental controls. Thesis, Department of Rangeland Ecology and Management, Texas A&M University, College Station, Texas. 102 p.

_____, S. ARCHER, T. W. BOUTTON, F. M. HONS AND D. A. ZUBERER. 2004. Soil respiration and nutrient cycling in wooded communities developing in grassland. Ecology, 85:2804-2817.

MCKINLEY, D. C. AND J. M. BLAiR. 2008. Woody plant encroachment by Juniperus virginiana in a mesic native grassland promotes rapid carbon and nitrogen accrual. Ecosystems, 11:454-168.

NORRIS, M. D., J. M. BIAIR AND L. C.JOHNSON. 2001. Land cover change in eastern Kansas: litter dynamics of closed-canopy eastern redcedar forests in tallgrass prairie. Can. J. Bot., 79:214-222.

OLSON, J. S. 1963. Energy storage and the balance of producers and decomposers in ecological systems. Ecology, 44:322-331.

PACALA, S. W., G. C. HURTT, D. BAKER, P. PEYUN, R. A. HOUGHTON, R. A. BIRDSEY, L. HEATH, E. T. SUNDQUIST, R. F. STALLARD, P. CIAIS, P. MOORCROIT, J. P. CASPERSEN, E. SHEVUAKOVA, B. MOORE, G. KOHLMAIER, E. A. HOLLAND, M. GLOOR, M. E. HARMON, S. M. FAN, J. L. SARMIENTO, C. L. GOODALE, D. S. SCHIMEL AND C. B. FIELD. 2001. Consistent land- and atmosphere-based U.S. carbon sink estimates. Science, 292:2316-2320.

POLLEY, H. W., B. J. WILSEYANDJ. D. DERNER. 2007. Dominant species constrain effects of species diversity on temporal variabilit in biomass production of tallgrass prairie. Oikos, 1 16:2044-2052.

REICH, P. B., D. W. PETERSON, D. WEDIN AND K. WRAGE. 2001. Fire and vegetation effects on productivity and nitrogen cycling across a forest-grassland continuum. Ecology, 82:1703-1719.

RYBA, S. A. AND R. M. BURGESS. 2002. Effects of sample preparation on the measurement of organic carbon, hydrogen, nitrogen, sulfur, and oxygen concentrations in marine sediments. Chemosphere, 48:139-147.

SAS. 1996. SAS for Windows, version 6.11. Gary, North Carolina.

SCHLESINGER, W. H. AND A. M. PILMANIS. 1998. Plant-soil interactions in deserts. Biogeochemistiy, 42:169-187.

_____, J. F. REYNOLDS, G. L. CUNNINGHAM, L. F. HUENNEKE, W. M. JARRELL, R. A. VIRGINIA AND W. G. WHITFORD. 1990. Biological feedbacks in global desertification. Science, 247:1043-1048.

SCURLOCK, J. M. O., K. JOHNSON AND R. J. OLSON. 2002. Estimating net primary productivity from grassland biomass dynamics measurements, dob. Change Biol., 8:736-753.

SMEINS, F. E. AND D. D. DIAMOND. 1983. Remnant grasslands of the Fayette Prairie, Texas. Am. Midi. Nat., 110:1-13.

SMITH, D. L. AND L. C.JOHNSON. 2003. Expansion oí Juniperus virginianaL. in the Great Plains: Changes in soil organic carbon dynamics. Glob. Biogeochem. Cy., 17:1062.

_____ AND _____. 2004. Vegetation-mediated changes in microclimate reduce soil respiration as woodlands expand into grasslands. Ecology, 85:3348-3361.

STEINAKER, D. F. AND S. D. WILSON. 2005. Belowground litter contributions to nitrogen cycling at a northern grassland-forest boundary. Ecology, 86:2825-2833.

VETAAS, O. R. 1992. Micro-site efiects of trees and shrubs in dry savannas. J. Veg. Set., 3:337-344.

WALDROP, M. P. AND M. K. FIRESTONE. 2004. Microbial community utilization of recalcitrant and simple carbon compounds: impact of oak-woodland plant communities. Oecologia, 138:275-284.

WESSMAN, C., S. ARCHER, L.JOHNSON AND G. ASNER. 2004. Woodland expansion in US grasslands: assessing land-cover change and biogeochemical impacts, p. 185-208. In: G. Gutman and M. A. Cochrane (eds.). Land Change Science: Observing, Monitoring and Understanding Trajectories of Change on the Earth's Surface. Kluwer Academic Publishers, New York.

WILSON, S. D. AND H. R. KLEB. 1996. The influence of prairie and forest vegetation on soil moisture and available nitrogen. Am. Midi. Nat., 136:222-231.

SUBMITTED 22 JUNE 2010 ACCEPTED 12 OCTOBER 2012

Author affiliation:

R. L. MCCULLEY*

Department of Biology, Duke University, Durham, North Camuña 27708

AND

R. B. JACKSON

Nicholas School of the Environment and Earth Sciences, Duke University, Durham, North Carolina 27708

Author affiliation:

1 Corresponding author's Present address: Department of Plant & Soil Sciences, N-222D Ag. Sci. North, University of Kentucky, Lexington 40546; Telephone: (859) 257-6388; FAX: (859) 323-1952; e-mail: rebecca.mcculley@uky.edu