This paper aims at understanding the effect that shrub encroachment into mesic mountain grasslands has on soil organic carbon (SOC) stocks. Thus we compared organic C and N contents in contiguous soil profiles under a conifer shrub, a legume shrub, and grasses (mesic grasslands) on 21 sites. Soil C and N recalcitrance indexes (RIC and RIN) were estimated as the ratio of unhydrolyzable C and N to total C or N. Contrary to our hypothesis, shrub encroachment did not cause a well-defined change in soil C stocks. Only a slight increase in C was detected in the top 15 cm soil layer under both shrubs. The C accretion rate in this soil layer was estimated to be approximately 28–42 g m−2 per annum. Legume shrub encroachment also produced a slight decrease in the C/N ratio in the top 15 cm soil layer. No significant changes in the RIC were detected as a result of encroachment; however, slight decreases in the RIN were found at medium depths. Both RIC and RIN showed a negative relationship with site temperature in the upper legume-shrub soil layers but not in conifer shrubs or under grasses, suggesting a priming effect on the recalcitrant soil stocks produced by the higher-quality inputs derived from legume encroachment.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 A generalized shift from primary to tertiary industries has taken place in most European mountain areas, where traditional and sustainable multifunctional activities have often been substituted by purely economically oriented ones [European Environment Agency, 1999]. This has resulted in an intensification of pasture use in the most accessible valleys and slopes and pasture abandonment in remote areas [Cernusca et al., 1996; Garcia-Ruiz et al., 1996]. Pasture abandonment can adversely affect species diversity, water storage capacity, soil quality, and other ecological and cultural goods and services [MacDonald et al., 2000]. Shrub expansion has been documented in grasslands in the Pyrenees [Molinillo et al., 1997; Roura-Pascual et al., 2005] and the Central System ranges [Sanz-Elorza et al., 2003] of the Iberian peninsula. This woody encroachment both increases the risk of fire propagation, by incrementing fuel load and fuel continuity, and reduces the habitat of some endangered species. Moreover, because mesic grasslands retain large amounts of soil C [Briggs et al., 2005; Garcia-Pausas et al., 2007] and woody proliferation into grasslands is a worldwide phenomenon [e.g., Archer et al., 1995; Fang et al., 2001], a better understanding of how shrub encroachment alters C dynamics is necessary for quantifying and balancing the global C budget [Lett et al., 2004].
 Following the Kyoto Protocol, the need for a better understanding of the processes and mechanisms leading to loss or sequestration of soil organic carbon (SOC) is universally recognized. Changes in land use are widely accepted as key drivers of global C dynamics [Guo and Gifford, 2002]. Nevertheless, despite the global significance of shrub proliferation, its biogeochemical consequences are poorly known, and the response of soil C to woody encroachment is uncertain [Asner et al., 2004]. A comparison of paired southwestern United States shrubland and grassland sites showed that SOC in shrublands was lower than in grasslands when the annual precipitation was above 600 mm [Jackson et al., 2002]. This decrease may be due to a change in C and N productivity and storage from belowground in open mesic grasslands to aboveground in shrub islands [Lett et al., 2004]. In contrast, other studies reported either no net C change or increased C with woody encroachment into mesic grasslands [Hibbard et al., 2003; Smith and Johnson, 2003].
 Shrub expansion into grasslands alters many aspects of the physical and biological environment. The lower soil respiration in the woodland relative to its paired grassland was related to the vegetation-mediated reduction in soil temperature under the canopy [Smith and Johnson, 2004]. In addition to microclimatic factors, the quality of aboveground and belowground C inputs may affect soil organic matter decomposition [Menteemeyer, 1978; Murphy et al., 1998; Knops et al., 2002]. Thus root chemistry has been described as the primary controller of root decomposition [Silver and Miya, 2001]. Soil organic matter is a complex mixture of compounds with different decomposition kinetics. Acid hydrolysis is a procedure commonly used to isolate stable organic matter [Six et al., 2002], and the amount isolated is considered to represent the recalcitrant fraction [Rovira and Vallejo, 2000; Tan et al., 2004b]. Differences in plant productivity and input quality between shrubs and grasses may affect the accumulation and decomposition patterns of soil organic matter pools. If shrubs have lower productivity, shrub encroachment is likely to result in net losses of C and N, especially belowground. However, the generally lower quality of shrub litter and roots (higher C to N ratio; lignified tissues, secondary compounds) may hinder soil organic matter decomposition and favor C accumulation [Paustian et al., 1997; Murphy et al., 1998; Moorhead et al., 1999]. On the other hand, the presumably higher N content of legume-shrub litter and roots suggests that decomposition may be greater in legume-dominated patches.
 Because our preliminary results indicated that abandoned subalpine grasslands in the Pyrenees had less SOC than grazed ones [Casals et al., 2004], we hypothesized that SOC stocks would be lower under shrubs than under contiguous mesic grasslands. Moreover, as the quality of soil organic matter drives SOC dynamics, we additionally hypothesized that the reduction in SOC content would be greater under legume shrubs than under conifer shrubs or grasslands. This study thus aimed at understanding (1) how shrub encroachment into mesic mountain grasslands affects SOC stocks and (2) how the observed effects depend on the quality of plant inputs. Thus in supraforest landscapes dominated by mesic grasslands we compared OC stocks in contiguous soil profiles under a conifer shrub (juniper, Juniperus communis L.), a legume shrub (broom, Cytisus balansae ssp. europaeus (G. López and Jarvis) Muñoz Garmendia), and grasses (mesic grasslands).
2. Material and Methods
2.1. Experimental Design
 Our study was designed to assess the effects that shrub encroachment into mesic mountain grasslands have on SOC stocks; for this we compared the soil profiles under shrubs known to have developed on grasslands with those in remnant grasslands. Using historical aerial photographs and information provided by local foresters, we made sure that shrub invasion had taken place in the last 40–50 years (a) and had been continuous for at least 15 a. A subsequent estimation of shrub age by ring counting corroborated this. In addition, the presence of the nonresprouting J. communis shrub confirmed that fire had not been present in the area during this time frame.
 On each site we located three plots (1 m2) in contiguous areas with similar slope, aspect, and soil series: one in an open grassland, one under the legume shrub C. balansae, and the third one under the conifer shrub J. communis. On each plot we estimated the aboveground and belowground shrub biomass and opened an entire soil profile. This design, based on side-to-side comparisons, is constrained by space-for-time assumptions [Picket, 1989], but it is nonetheless a widely accepted technique in ecology [Kratz et al., 2003]. Thus we assumed the present-day stocks under the grassland to be an acceptable baseline for assessing the differences derived from shrub encroachment. To minimize the side-to-side differences, the plots were located as close to each other as possible and never more than 10 m apart. In this way, no differences in soil texture or soil depth were found between the profiles on the same site. Few differences in soil texture existed between localities and soil pH ranged from 4 to 5 (Table 1). Because our design aimed at maximizing the differences generated by shrub encroachment, to select the plot we looked for the shrub with the largest stem, which we assumed to be the oldest.
Table 1. Physiographic and Climatic Characteristics of the Sites Studieda
 The research was conducted on 21 sites widely distributed in the Pyrenees and Central System mountains of the Iberian peninsula. We selected 16 sites in the Pyrenees at the subalpine stage, from 1700 to 2100 m above sea level (asl), and five sites in the Central System at the oromediterranean stage, from 1650 to 2050 m asl (Table 1).
 In the Pyrenees the mesic grasslands were dominated by Festuca nigrescens Lam., Agrostisrupestris All., and Festuca eskia Ramond ex DC. The Guadarrama oromediterranean grasslands were dominated by Festuca violacea Gaud. var. iberica Hack., Festuca ovina L. spp. heterophylla (Lamk.) Hack., and Deschampsia flexuosa (L.) Trin. ssp iberica Rivas Mart.
 We chose localities in which conifer (J. communis) and legume (C. balansae) shrubs coexisted with mountain grasslands. In general, C. balansae grows on south facing aspects. Thus most of our sites were situated on intermediate slopes with this slope orientation. The lithology was either metamorphic (slates or shales) or plutonic (granodiorite). Climatic variables were estimated using a model that combines climatic data and topographic variables into a GIS system [Ninyerola et al., 2000]. Mean annual temperature and precipitation values ranged from 3 to 7 °C and from 700 to 1250 mm, respectively (Table 1).
2.3. Aboveground and Belowground Biomass
 Aboveground shrub biomass was estimated by cutting and weighing all the aerial shrub biomass in a 1 × 1 m area, taking the shrub base as the center. A wood slice from the largest basal stem was collected to estimate shrub age by ring counting. Aboveground grassland biomass was not quantified. For general comparisons, data from ongoing studies in the Pyrenees were used (P. Casals, unpublished data 2006). Fine roots in the top 15 cm soil layer were estimated by taking a soil core (5 × 5 cm, 15 cm long) in both shrubs and grasses. In the laboratory, roots were separated from the soil by flotation, oven dried (60°C), and weighed. In ten localities, legume and conifer shrub woody root biomass was determined in a 0.5 × 0.5 × 0.5 m soil cube by sieving and handpicking medium and coarse roots (diameter > 2.5 cm). Fresh aboveground and belowground biomasses were corrected by their respective fresh/dry weight ratio obtained from subsamples collected and dried (60°C) in the laboratory to constant weight. We collected the litter mass from the center of each shrub plot using a metallic cylinder with a 490 cm2 effective area. The litter was dried to constant weight (60°C).
2.4. Soil Sampling
 On each plot we dug a soil profile to the parent rock. We described soil morphology and collected volumetric soil samples for several depth increments (0–5 cm, 5–15 cm, 15–30 cm, 30–50 cm, 50–70 cm, 70–90 cm, and 90–110 cm). At each profile depth we estimated soil bulk density. The fine-earth fraction was calculated after subtracting the proportions of stone and gravel content. Stoniness was estimated in the field by weighing the stones sampled at each depth. In the laboratory the fine earth was weighed after sieving (<2 mm), and the gravel content was calculated.
2.5. Carbon and Nitrogen Pools
 For each plot and site a subsample of aboveground and belowground shrub biomass, litter, and soil was ground up in a ball mill (MM200 Retsch®). The C concentration in the aboveground and belowground biomass and in the litter was determined by means of an elemental analyzer (Carlo Erba NA 1500 analyzer), while the C concentration in the soil was determined by dichromate oxidation in an aluminum heating block [Nelson and Sommers, 1996]. The total N concentration in the biomass, litter, and soil of all three plant types was also determined using the elemental autoanalyzer. The total SOC or N profile was estimated by integrating the element's concentration in each layer, corrected by its thickness, soil bulk density, stoniness, and gravel content.
 Soil recalcitrant C and N were obtained by acid hydrolysis [Rovira and Vallejo, 2000]. Briefly, between 600 and 800 mg of ground soil samples were hydrolyzed with 25 mL of 6M HCl in sealed Pyrex tubes at 105°C for 18 h, with occasional shaking, using an aluminum heating block. After cooling the unhydrolyzed residue was recovered by repeated centrifugation and decantation of the supernatant liquid using deionized water to eliminate residual HCl. The residue was transferred to preweighed vials, dried at 60°C to constant weight, and analyzed for carbon and nitrogen with the Carlo Erba analyzer.
 All the data were presented as mean ± standard error for the 21 sites (n = 21). We did side-by-side comparisons of C and N stocks on three contiguous plots with different plants: grasses, legume shrub, and conifer shrub. This approach was chosen to minimize the confounding effects of climate and soil type between the 21 sites. We tested C and N differences in shrub biomass (aboveground and belowground) and litter using a paired t test. We assessed differences between grasses, legume, and conifer with respect to fine-root biomass and soil C and N stocks (total or recalcitrant fractions) using a General Linear Model (GLM)–Repeated Measures analysis with the three plant types as levels of a within-subject factor (plant type). Pairwise modified Bonferroni means comparisons were performed on the significant main effect from the repeated measures tests [Day and Quin, 1989]. All effects were considered statistically significant when p < 0.10. Since the observed trends in soil stocks and the differences between plant types could depend on the mean annual site temperature, this parameter was added to the model as a covariable. When the interaction between plant type and temperature was significant, a posteriori multiple comparisons were carried out by the simple contrast. This post hoc contrast compares a selected level of the within-factor (conifer or legume shrub) to a reference level (grasses). All statistical analyses were performed using SPSS version 11.
3.1. Shrub C and N Pools
 Aboveground biomass of conifer and legume shrubs ranged from 2.59 kg m−2 to 9.11 kg m−2 (Table 2). Medium and coarse shrub root biomass in the top 50 cm was larger in the legume shrub than in the conifer one. Fine roots in the top 15 cm soil layer did not differ between the three plant types considered (2.78 ± 0.43 kg m−2). No differences were detected between shrubs with respect to aboveground or belowground C concentrations, while, as expected, N concentrations in both aboveground and belowground biomass were higher in legumes than in conifers (2.38 ± 0.17% and 1.87 ± 0.17% in aboveground and belowground biomass of legumes versus 1.46 ± 0.13% and 0.76 ± 0.08% in aboveground and belowground biomass of conifers). As a result, the C to N ratio was significantly lower in the legume biomass. No differences were found between shrubs with respect to litter mass or litter C concentrations. Again, N concentration was higher in legume litter than in conifer litter (2.74 ± 0.22% and 1.69 ± 0.06%, respectively).
Table 2. Aboveground and Belowground Shrub and Litter Biomassa
C biomass and C to N ratio on conifer and legume shrubs (mean ± SE; n = 21). Significance levels of t tests are indicated as p values.
Biomass, kg m−2
5.31 ± 0.47
6.36 ± 0.95
3.77 ± 0.70
2.73 ± 0.57
2.63 ± 0.30
2.67 ± 0.32
C Biomass, kg m−2
2.73 ± 0.24
3.24 ± 0.48
1.85 ± 0.34
1.35 ± 0.28
1.17 ± 0.13
1.16 ± 0.14
C to N ratio
22.1 ± 1.9
36.1 ± 3.6
27.0 ± 2.3
68.4 ± 8.6
16.1 ± 0.5
26.0 ± 1.0
3.2. Soil Organic C and Total N Stocks
 In the entire soil profile, SOC ranged from 4.45 kg m−2 to 32.09 kg m−2, and N ranged from 0.51 kg m−2 to 2.95 kg m−2. No differences in SOC were observed between the three plant types (15.55 ± 1.55 kg m−2). In contrast, total N in the entire soil profile of legume shrubs was higher than in the grassland profiles (1.54 ± 0.13 kg m−2 versus 1.38 ± 0.13 kg m−2), while total N in conifer profiles (1.44 ± 0.14 kg m−2) was not different from those of legume shrubs or grasslands.
 By depths both shrub types had higher organic C concentrations in the 5–15 cm layer than grasslands (Figure 1a), and considering the first 15 cm of soil as a whole, both shrubs also had higher SOC than grasslands (Table 3). Moreover, total N in the top 15 cm soil layer was higher under both shrubs than in grasslands (Table 3). The soil C/N ratio in the first 15 cm of soil was slightly lower under legume shrubs than under the other plant types (Figure 1b).
Table 3. Total and Recalcitrant Organic C and Total N in the Fine-Earth Fraction of Soil Horizons Under Three Plant Types (Conifer and Legume Shrubs and Grasslands)a
Total Soil Stocks, kg m−2
Recalcitrant Fraction, kg m−2
Units for C and N are kg m−2. Mean ± SE; n = 21. For each pool and horizon, significances of repeated measures analysis (within-factor effects) are indicated (p value). When significant (p < 0.10), different letters (a and b) represent mean differences as determined by pairwise-modified Bonferroni mean comparison test.
9.49 ± 0.74 b
10.07 ± 0.87 a
10.31 ± 0.88 a
4.79 ± 0.38 b
5.20 ± 0.48 ab
5.48 ± 0.48 a
7.08 ± 0.87
7.77 ± 1.10
7.39 ± 0.74
3.79 ± 0.45
4.14 ± 0.68
4.02 ± 0.45
7.78 ± 1.12
8.97 ± 1.42
8.15 ± 1.33
3.86 ± 0.58
4.79 ± 0.80
4.16 ± 0.69
0.86 ± 0.06 b
0.99 ± 0.07 a
0.96 ± 0.07 a
0.15 ± 0.02
0.17 ± 0.02
0.16 ± 0.01
0.68 ± 0.07 b
0.77 ± 0.08 a
0.73 ± 0.07 ab
0.17 ± 0.02
0.17 ± 0.02
0.17 ± 0.02
0.72 ± 0.11
0.81 ± 0.11
0.75 ± 0.10
0.22 ± 0.03
0.24 ± 0.03
0.23 ± 0.03
3.3. Soil Recalcitrant C and N Stocks
 The recalcitrant C fraction (RC) in the top 15 cm soil layer was higher under conifers than in grasslands, with the legume value being intermediate between them (Table 3). We did not detect any differences between plant types with respect to either the RC of the remaining soil layers or the recalcitrant N (RN) of the first 50 cm of soil (Table 3).
 The RIC ratio stayed around 50% and remained almost constant with depth, with no differences between plant types (Figure 2a). In contrast, the RIN ratio clearly increased with depth and with lower values under legume shrubs in the 15–30 cm soil depth.
3.4. Soil Stocks and Shrub Ages
 In general, the conifer shrubs were about 13 a older than the legume shrubs (32 ± 2 a old versus 19 ± 1 a old in conifer and legume, respectively; paired t student test, p < 0.001, n = 21). Linear regressions showed that the age of each shrub was positively related with the differences in SOC or N calculated between their respective profiles and the grassland ones in the 0–15 cm soil layer (Table 4). Since at zero time (shrub age = 0) all shrublands were grasslands, their initial stocks were the same, and the constant term in the equation was assumed to be zero. The regressions in the remaining soil layers were not significant. Assuming both that there is a constant accumulation rate and that the current SOC under grasslands is an acceptable baseline for C accretion estimates, the regression coefficient indicated a C accretion rate of about 28 to 42 g m−2 per annum in the first 15 cm of soil (independent variable coefficient, Table 4). Comparison of standard errors showed that legume and conifer C stocks regression coefficients did not differ, while the coefficient for total N was higher for legume shrubs than for conifer shrubs (Table 4).
Table 4. Linear Regressions Between the Differences in Litter or Soil Stocks in the First 15 cm of Soil Under Each Shrub and Their Paired Grassland Against Shrub Agea
The constant term (a, in the equation, y = a + bx) is assumed to be zero, because at zero time (age = 0) all shrublands were grasslands, and their stocks were assumed to be equal to those of the adjacent grasslands. For the slope of the equation (b) the estimated value ± standard error of the estimation is given. The significance of the obtained r value is given as p values. For linear regressions, n = 21; units for differences are g m−2, and units for shrub age are years.
56.52 ± 7.79
33.27 ± 4.43
Soil organic C
42.32 ± 13.87
28.05 ± 8.91
Soil recalcitrant C
3.26 ± 1.19
2.48 ± 0.81
3.53 ± 0.48
1.32 ± 0.18
Soil total N
6.56 ± 1.49
2.77 ± 0.75
Soil recalcitrant N
0.11 ± 0.05
0.04 ± 0.02
 Using the yearly stock changes estimated by these regressions (Table 4) for each locality, SOC, total N, and the recalcitrant pools in the 0–15 cm soil layer were estimated to 25 a of shrub encroachment (SOC25, N25, RC25, and RN25), which is the intermediate age between both shrub mean ages. SOC25 under legume shrubs continued to be higher than under grasslands (10.27 ± 0.85 kg m−2 versus 9.49 ± 0.74 kg m−2, respectively), while SOC25 under conifer shrubs (10.16 ± 0.87 kg m−2) was not different from the legume or grassland values. N25 in the top 15 cm of legume soil was higher than under conifers or grasslands (1.03 ± 0.07 versus 0.94 ± 0.07 and 0.86 ± 0.06 for legume, conifer, and grasslands, respectively).
 Litter C accumulation rate was higher under legume shrubs (56.52 ± 7.79 g m−2 per annum) than under conifer shrubs (33.27 ± 4.43 g m−2 per annum) (Table 4). Litter N accumulation rate was also higher under legume shrubs than under conifer shrubs (Table 4).
3.5. Soil Organic C and Total N Stocks and Site Characteristics
 The variability in soil stocks was correlated with site factors such as profile depth, gravel content, macroclimate, and altitude. Soil depth explained about 20% of the variability in SOC or N. SOC and total N in the 0–50 cm soil profile correlated negatively with temperature and positively with precipitation (Table 5).
Table 5. Pearson Correlation Coefficients Between Organic C and Total N in the Fine-Earth Fraction of the Top 50 cm of Soil and Other Selected Site Variablesa
Mean site soil stocks from the three plant types were used; n = 21. The significance of the obtained r values is given as p values.
Soil organic C
 Site temperature correlated negatively with SOC25 in the first 15 cm of soil without interaction with the plant type (Figure 3a). In the 15 to 30 cm layer the SOC/site temperature trend was maintained only in the legume profiles (Figure 3b). N25 in the first 15 cm and N in the other horizons also showed negative relationships with site temperature (Figures 3d, 3e, and 3f), without interactions between plant type and site temperature.
 The above relationships (positive with precipitation and negative with temperature) were also observed for the recalcitrant (nonhydrolyzable) pool (data not shown). The recalcitrant to total stocks ratios (RIC and RIN) did not show any pattern against temperature in grasslands or conifer soils; however, under legume shrubs, RIC and RIN in the first 30 cm correlated negatively with temperature (Figure 4).
4.1. C Stocks in Grasslands and Shrubs
 Soil organic carbon (SOC) stocks in subalpine and alpine grasslands in the Pyrenees range from 6.5 to 30 kg C m−2 [Garcia-Pausas et al., 2007]. In our study, SOC content under grasses and shrubs ranged from 6 to 32 kg C m−2. Overall, total OC content in the soil profiles was about 16 kg m−2, with no paired differences between grasslands and shrubs. Although the effect of herbivory on plant productivity and C allocation is still under debate, abandonment of light, extensive grazing management might reduce soil stocks [Schuman et al., 1999; Pucheta et al., 2004]. Thus in a survey of grasslands in the Pyrenees, Casals et al.  found that abandoned grasslands had lower OC stocks in the first 20 cm of soil than grazed ones. Other authors [Jackson et al., 2002] postulated that woody encroachment into mesic grasslands might produce a net loss of SOC. Nevertheless, our results showed that the SOC under both shrubs was not lower than under the adjacent mesic mountain grasslands. The reasons for these contradictory findings are not clear. On the one hand, they may be related to differences between the observed shrub species with respect to their allocation patterns or their mediated changes in microclimate, as suggested by Smith and Johnson . On the other hand, as neither study controlled for previous grazing regimes, it is possible that the distinct, unknown, grazing histories at each site may have contributed to the contradictory results.
 In the 15–60 a old shrubs considered in our study, the C content in shrub biomass and litter was found to be about 5 to 7 kg m−2. Although we did not estimate the C content in the aboveground biomass of our grasslands, an extensive survey in subalpine grasslands in the Pyrenees reported that the C content in aboveground biomass was lower than 0.3 kg m−2 (M. T. Sebastià, personal communication, 2006). As no differences were found in fine-root biomass between plant types, woody encroachment into subalpine mesic grasslands should drive an increase in C stocks of about 4 to 6 kg m−2, compared with remnant grasslands, mainly as a consequence of C stocks in shrub biomass and litter. These results contradict those found by Jackson et al. , who observed losses in the net balance of C on two of three mesic sites after woody encroachment. However, as Goodale and Davidson  pointed out, Jackson et al.  did not measure the C stocks in woody roots (those of 1 cm or more in diameter), and this could have affected their calculations of the net balance of C.
 Shrub encroachment into grasslands has been considered a potential way to store OC deep in the soil, because of the deeper roots of shrubs [Jobbágy and Jackson, 2000]. Nevertheless, in our study, as in others [Hibbard et al., 2001; Smith and Johnson, 2003], it was found that OC concentrated in the shallow soil layers after woody encroachment. Thus shrub encroachment into grasslands increased OC by about 6–8%,and total N increased by about 10–15% in the top 15 cm of mineral soil. These differences may be due either to increased C inputs from shrub litter and roots or to reduced soil C decomposition from the less favorable microclimate for decomposition found under shrubs (as described for mesic grasslands [McCarron et al., 2003; Smith and Johnson, 2004]). An additional reason, a reduction in substrate quality after woody encroachment, is less likely, as we discuss in section 4.2. When the top 0 to 50 cm of soil were integrated, the differences in SOC between shrubs and grasslands disappeared.
 Assuming that present-day soil stocks under grasslands are an acceptable baseline for comparing C accretion in the top 15 cm soil layer under both types of shrubs and that C inputs were linear along the woody encroachment span, the C accretion ratio after shrub encroachment was about 28–42 g m−2 per annum. These amounts are similar to the accretions described in a subtropical savanna ecosystem (from 8 to 23 g m−2 per annum [Hibbard et al., 2001]) or after abandonment of agricultural land in temperate regions (28 g m−2 per annum [Schlesinger, 1997]).
4.2. Plant Quality Effects on Soil C and N Stocks
 The presence of different-aged shrubs on the same site could be explained by differences in colonization strategies: Conifer shrubs rely on animals and legume shrubs are based in either clonal growth or explosive dehiscence of fruits. The different shrub ages may partially confound the soil stock comparisons between legume and conifer shrubs. Thus after correcting the soil accretion age to 25 (i.e., the intermediate age between both shrub mean ages), the differences between grasslands and shrubs with respect to the soil stocks in the top 15 cm soil level could be found only under legume soils. In general, higher substrate quality stimulates organic matter decomposition, and the C to N ratio (regardless of its simplification) is a valuable predictive tool in relation to the decomposition of both litter [Menteemeyer, 1978; Moro and Domingo, 2000; Shaw and Harte, 2001] and roots [Silver and Miya, 2001]. In our study, although conifer shrubs showed lower N content in their aboveground and belowground biomass and litter than legume shrubs, their soil C and N stocks were similar (or slightly lower after age correction to 25 a because of soil encroachment). Soil C accretion after N fixer colonization or plantation have also been described [Kaye et al., 2000; Resh et al., 2002]. A combination of two different processes has been argued by these authors: greater N fixer C inputs to the soil and/or inhibition of humified soil C decomposition due to increased N availability. Our study supports the first explanation. The greater medium and coarse root biomass in the first 50 cm of soil under legume shrubs suggested higher C inputs to the soil from root decomposition.
 In contrast, the results obtained in recalcitrance analysis suggest that encroachment did not involve major changes in the biochemical quality of the soil organic matter. The unhydrolyzable organic matter is considered to represent the recalcitrant fraction [Rovira and Vallejo, 2002; Tan et al., 2004b], and its amount has been quoted as a measure of the nonactive OM pool [Six et al., 2002; Paul et al., 2006]. Small increases in RIC values were detected (Figure 2a), but they never reached significance. This result was somewhat unexpected because shrub proliferation (at least for conifer shrubs) should result in both an increase in the inputs of lignified plant debris and an overall decrease in the quality of organic inputs into the soil. As for changes in RIN, when detected, they were in the sense of higher amounts of labile nitrogen in the soil under shrubs, relative to the original grassland soil (Figure 2b). Increased root exudates, or increased leaching of labile nitrogenous compounds from litter, could explain this observation which deserves further research.
 It is worth noting that, in the soils studied in this paper, the distribution of RIC and RIN values along the profile differed greatly from what we had previously observed for forest soils [Rovira and Vallejo, 2007]. While in the forest soils studied the RIC values clearly tended to decrease with depth (being highest in H horizons and lowest in B horizons), in the grass and shrub soils analyzed the quality of the carbon clearly stayed more or less constant throughout the profile, perhaps reflecting a homogeneous quality of the organic inputs in all horizons. On the other hand, the observed increase in RIN with depth could reflect an N-aging process, since N is expected to be more biochemically evolved in deep layers than in upper horizons. In a previous work [Rovira and Vallejo, 2002] we reported that N recalcitrance increased with decomposition in plant debris incubated into the soil. Nevertheless, in forest soils we observed an overall maintenance of RIN values with depth, and we detected clear increases in RIN values only in deep horizons that were very poor in N [Rovira and Vallejo, 2007].
4.3. Effects of Temperature and Precipitation on Soil C and N Stocks
 At regional scale a positive response of soil C to precipitation has been related to an increase in either aboveground production or C inputs to the soil [Paruelo et al., 1998], while a negative response of soil C to temperature has been associated with the positive effect this climatic variable has on decomposition [McDaniel and Munn, 1985]. In regions where water availability is not a limiting factor, both production and decomposition increase with temperature, but relative increases in decomposition are greater [Schlesinger, 1997; Oades, 1988; Kirschbaum, 2000; Tan et al., 2004a]. An inverse relationship between SOC and temperature has been reported in a number of gradient studies in grasslands at regional scales [Burke et al., 1989; Paruelo et al., 1998]. Our results agree with this trend. Thus both SOC and N were negatively correlated with site mean temperature and positively correlated with annual precipitation and altitude.
 The ability to predict C feedback to global change requires a mechanistic understanding of how C cycling is affected by both climate and species composition [Kueppers and Harte, 2005]. The negative relationship found between soil stocks and temperature in our study suggests that an increase in temperature might represent a loss in both total and recalcitrant SOC and N stocks. Contradictory results have recently appeared in the bibliography on whether total organic matter decomposition and recalcitrant pools decomposition react in the same way to increased temperatures. While Fang et al.  concluded that no detectable differences exist, Knorr et al.  found that the decomposition of recalcitrant organic matter is more sensitive to increased temperatures than that of SOC as a whole, and Bol et al.  observed that high temperatures tend to favor the decomposition of the recalcitrant SOM pool. Our results showed a decrease in both RIC and RIN with increasing temperatures in legume upper soil layers, but not in either conifer or grassland profiles. The hypothesis that the higher quality of legume inputs might exert a priming effect and stimulate the relative decomposition of recalcitrant stocks with respect to total soil C decomposition requires further research.
 This study shows that woody encroachment did not decrease soil C stocks in mesic mountain grasslands. In fact, as a result of the higher amounts of shrub aboveground and belowground biomass and litter, the C stocks in shrubs were about 4 to 6 kg m−2 higher than in grasslands. This C is accumulated mainly aboveground and is thus more highly subjected to natural or anthropogenic perturbations than soil C. Although no differences existed between shrubs and grasses in the SOC stocks of the entire soil profile, woody encroachment slightly increased soil C and N stocks in the top 15 cm soil level. As the stocks under both shrubs were similar, the presumably greater decomposition under legumes as a consequence of their higher N content might be compensated by higher belowground productivity. Additional research is required to elucidate the role of substrate quality inputs on soil organic fractions dynamics.
 This research was partially funded by both the Spanish Research Agency (MEC, CGL2005-08133-CO3) and the Spanish Ministry of Environment. The first author (F. M.) has a Ph.D. fellowship from DURSI-Generalitat de Catalunya and the European Social Fund (2005FI 00801). We thank all the technicians that gave us support in the localization of the sites. J. Juery, P. Issert, M. Taull, and J. C. Loaiza assisted in the field. Contributions by J. Retana, J. Romanyà, J. Garcia, S. Letcher, and E. Marks improved early manuscript drafts. CEAM is supported financially by the Generalitat Valenciana and Bancaixa.