Small-Scale Forest Structure Influences Spatial Variability of Belowground Carbon Fluxes in a Mature Mediterranean Beech Forest

The tree belowground compartment, especially fine roots, plays a relevant role in the forest ecosystem carbon (C) cycle, contributing largely to soil CO2 efflux (SR) and to net primary production (NPP). Beyond the well-known role of environmental drivers on fine root production (FRP) and SR, other determinants such as forest structure are still poorly understood. We investigated spatial variability of FRP, SR, forest structural traits, and their reciprocal interactions in a mature beech forest in the Mediterranean mountains. In the year of study, FRP resulted in the main component of NPP and explained about 70% of spatial variability of SR. Moreover, FRP was strictly driven by leaf area index (LAI) and soil water content (SWC). These results suggest a framework of close interactions between structural and functional forest features at the local scale to optimize C source–sink relationships under climate variability in a Mediterranean mature beech forest.


Introduction
Terrestrial ecosystems, especially forests, have an active role in the global carbon (C) cycle: forests cover about 4.2 × 10 3 Mha of the earth's land surface, accounting for about 45% of terrestrial carbon and contributing to about 50% of terrestrial net primary production (NPP) [1]. As we are following the climatic scenario characterized by the highest variations [2], forests play a crucial role to mitigate global climatic change by removing 2.4 ± 0.4 Pg C y −1 from the atmosphere through growth [3]. This amount corresponds up to 30% of anthropogenic CO 2 emissions from fossil fuel burning and deforestation [4]; hence, it is evident how changes in the productivity of the forest ecosystem affects the C-cycle.
Overall, the forest net effect on the carbon cycle is strictly related to net primary production (NPP), which is the small difference between the amount of C absorbed through photosynthesis and the C emitted by plant (autotrophic) respiration [5]. NPP is usually estimated as the new organic matter produced during a given period (generally one year), in both aboveground and belowground plant compartments, and it is affected by environmental drivers [6].
Despite the importance of fine roots in the ecological processes, our understanding on their dynamics is still limited [13,14]. Studies were mainly focused on the role of environmental drivers on FRP [15,16], evidencing the impact of FRP on SR, both within and among ecosystems [11,17,18]. Therefore, a better identification of drivers regulating belowground processes through FRP is essential for a correct estimation of the ecosystem C budget [19].
The forest structure, here defined as the distribution of trees over an area, is determined by past management practices and represents one of the major drivers of the forest C cycle [20][21][22] and biodiversity [23]. Indeed, the forest structure interacts with tree physiological functionality [13,24] and climate [19], affecting C allocation and source-sink relationships [25,26].
In this context, the general objective of this study was to explore the intra-site relationships among forest structures (number of trees, basal area, maximum diameter, and leaf area index), soil characteristics, and spatial variability of SR and FRP in a mature beech stand in Mediterranean montane conditions, characterized by an almost total canopy closure. Hence, FRP, SR, soil proprieties, and forest structural parameters were measured in different randomized plots inside the stand.
Specific aims of the present study were to assess (i) the fraction of annual NPP partitioned to FRP; (ii) if, and which, intra-stand forest structural parameters and soil characteristics affect FRP; and (iii) the effect of FRP on the spatial variability of SR.

Site Characteristics
The experiment was carried out during 2007-2008 in a European beech (Fagus sylvatica L.) forest near Collelongo (Abruzzi Region, Central Italy, Figure 1A), where a permanent experimental facility (Selva Piana stand, 41 • 50 58" N, 13 • 35 17" E, 1560 m elevation) was established in 1991. The Selva Piana stand is located within a 3000 ha community forest that is part of a wider forest area, included in the external belt of the Abruzzi National Park. The environmental and structural conditions of the stand are representative of central Apennine beech forests. In 2007, the stand density was 825 trees ha −1 , and the basal area was 40.5 m 2 ha −1 with a mean diameter at breast height of 25 cm and a mean height of 21.5 m. Mean tree age in 2007 was estimated to be about 115 years. The forest structure is characterized by a sensible vertical stratification derived from a conversion of a beech coppice with standards to high stand [24,27] started after the middle of the 20th century.
The soil, developed on calcareous bedrock, has a variable depth (40-100 cm) and is classified as a humic Alisol [28]. Site topography is gently sloping. The climate is Mediterranean montane, with a mean annual temperature of 6.97 °C, and the mean temperatures of the coldest and warmest months are −1.04 and 16.3 °C, respectively (average of 1996-2014). Mean annual precipitation is 1116 mm, of which ~10% falls in summer. During the study, in 2007, the summer was extremely dry with only 3 The forest structure is characterized by a sensible vertical stratification derived from a conversion of a beech coppice with standards to high stand [24,27] started after the middle of the 20th century.
The soil, developed on calcareous bedrock, has a variable depth (40-100 cm) and is classified as a humic Alisol [28]. Site topography is gently sloping. The climate is Mediterranean montane, with a mean annual temperature of 6.97 • C, and the mean temperatures of the coldest and warmest months are −1.04 and 16.3 • C, respectively (average of 1996-2014). Mean annual precipitation is 1116 mm, of which~10% falls in summer. During the study, in 2007, the summer was extremely dry with only 3 mm precipitation in July and August ( Figure 2).

Experimental Design
To assess the role of local forest structure on the studied parameters, nine circular and relatively small experimental plots (5 m radius) were randomly established, maintaining a minimum distance of 15 m between plots centers ( Figure 1B). Inside each plot, we measured forest structural parameters, soil CO2 efflux, FRP, and soil characteristics ( Figure 1C).

Fine Root Production (FRP)
The ingrowth core method (Ostonen et al. 2005) was used to estimate FRP. In each experimental plot, two ingrowth cores were installed at the beginning of April 2007. Cores were made of 0.4 cm mesh net of plastic material, to allow ingrowth of fine to medium roots (1 to 4 mm). The cores were cylindrical, with a base diameter of 5.5 cm and exploring a depth of 30 cm, as over 90% of fine roots are located at this soil depth [29][30][31]. Cores were filled using soil collected in the same stand near the experimental plots, air dried, and sieved at 0.4 mm to remove all the roots. One of the two cores was extracted 6 months later at the end of the growing season (October 2007), while the second core was collected one year after the installation, before the bud break (May 2008), because diffuse, porous ring species tend to produce a greater proportion of their roots after bud break [14]. After the extractions, cores were carried to the laboratory for fine roots collection (<2 mm). Hence, FRP for both annual (FRPY) and growing season (FRPG) scales was estimated. Finally, FRP for the leafless period (FRPLP, related to winter and early spring FRP) was calculated as the difference between FRPY and FRPG. In addition, other 11 plots were established, where only FRP was measured according to the abovedescribed protocol. This additional dataset was used only to increase to 20 the sampling points used for NPP estimation.

Experimental Design
To assess the role of local forest structure on the studied parameters, nine circular and relatively small experimental plots (5 m radius) were randomly established, maintaining a minimum distance of 15 m between plots centers ( Figure 1B). Inside each plot, we measured forest structural parameters, soil CO 2 efflux, FRP, and soil characteristics ( Figure 1C).

Fine Root Production (FRP)
The ingrowth core method (Ostonen et al. 2005) was used to estimate FRP. In each experimental plot, two ingrowth cores were installed at the beginning of April 2007. Cores were made of 0.4 cm mesh net of plastic material, to allow ingrowth of fine to medium roots (1 to 4 mm). The cores were cylindrical, with a base diameter of 5.5 cm and exploring a depth of 30 cm, as over 90% of fine roots are located at this soil depth [29][30][31]. Cores were filled using soil collected in the same stand near the experimental plots, air dried, and sieved at 0.4 mm to remove all the roots. One of the two cores was extracted 6 months later at the end of the growing season (October 2007), while the second core was collected one year after the installation, before the bud break (May 2008), because diffuse, porous ring species tend to produce a greater proportion of their roots after bud break [14]. After the extractions, cores were carried to the laboratory for fine roots collection (<2 mm). Hence, FRP for both annual (FRP Y ) and growing season (FRP G ) scales was estimated. Finally, FRP for the leafless period (FRP LP , related to winter and early spring FRP) was calculated as the difference between FRP Y and FRP G . In addition, other 11 plots were established, where only FRP was measured according to the above-described protocol. This additional dataset was used only to increase to 20 the sampling points used for NPP estimation.

Forest Structural Parameters
At the center of each sampling plot, leaf area index (LAI) was measured at the seasonal peak of 2007 (July) through two LAI 2000 Canopy Analyzers (Li-Cor) measuring above and below the canopy, respectively. LAI values were calculated using the software FV2200 (LICOR Biosciences, Lincoln, NE, USA) by considering only four of the five measuring rings to restrict the angle of view to better represent LAI of the sampling plots.
At the end of the experiment (May 2008), the diameter at breast height (DBH) of each tree inside the plots was measured, and the basal area (BA), representing the area (m 2 ) of the cross-section of the stem measured at 1.30 m height, mean, and maximum tree diameter of the plot (D max ) were derived.

Soil CO 2 Efflux, Microclimatic Condition, and Soil Characteristics
Inside the nine experimental plots, five PVC collars (10 cm diameter and 5 cm high, for a total of 45 points) were inserted in the soil with a circular distribution spaced at a minimum of 50 cm away from the neighboring trees ( Figure 1C Soil temperature (T Soil) and soil water content (SWC) were measured at 0-10 cm by means of STP-1 (PP-System, Hitchin, UK) and time domain reflectometry techniques (Trime-FM, IMKO, gmbH, Ettlingen, Germany), respectively. All measurements were performed concurrently to the SR sampling.
In May 2008, litter and soil samples, down to 30 cm depth, were collected inside and below each PVC collar installed for SR measurements.

Carbon/Nitrogen Concentration
Litter, soil, and fine root C and N content were determined by an elemental analyzer (Model NA 1500, Carlo Erba, Milan, Italy). Soil samples were previously treated with HCl (10%) to remove carbonates.

Statistical Analysis
Analysis of the relationships between fine root production with both forest structural parameters and soil characteristics were performed only on annual values of FRP (FRP Y ) because soil sampling and DBH measurements to calculate the forest structural parameters were carried out at the end of the experimental period in May 2008.
Stepwise analysis was used to select the independent variables determining FRP Y (Table 1 and  Table S1). We tested data normality and constant variance using the Shapiro-Wilk test and the Spearman rank correlation between the absolute values of the residuals and the observed value of the dependent variable. We applied communality analysis (CA) to a multiple linear regression built with the variables identified by the stepwise analysis to disentangle the effects of each independent variable. Communality analysis shares the explained variance into pure and joint effects of predictors in order to assess the relative contribution of each predictor to the explained variance of the response variable [32]. Table 1. Descriptive statistics of forest structural and soil parameters used to assess the relation with the FRP in the 9 sampling plots. Maximum (Max), minimum (Min), mean (Mean), and coefficient of variation (CV) values are reported for each parameter. N tree plot-1 is the number of tree inside each sampling plot; basal area (m 2 ) is the sum of the stem cross-section areas of the n trees present in each plot; D max (cm) is the maximum diameter measured in the sampling plot tree height; LAI (m 2 m −2 ); T soil ( • C) is the average soil temperature measured during the SR campaigns (n = 11); SWC (%) is the annual average soil water content measured during the SR campaigns, except February 2008 because of the snow cover (n = 10); SMN (%) is the organic nitrogen percentage in the mineral soil layer; SMC (%) is the organic carbon percentage in the mineral soil layer; SON (%) is the organic nitrogen percentage in the organic soil layer; SOC (%) is the organic carbon percentage in the organic soil layer; litter amount (g DW m −2 ); litter N (%) and litter C (%) are the nitrogen and carbon percentages of the litter, respectively.

Fine Root Production and Its Drivers
Fine root production during the growing season (FRP G ) was estimated at 7.63 ± 2.02 Mg ha −1 , ranging from 5.47 to 10.93 Mg ha −1 , with a coefficient of variation (CV) of 0.26. In May 2008, 12 months after in-growth cores installation, FRP Y was 9.80 ± 1.97 Mg ha −1 y −1 , ranging from 7.32 to 13.50 Mg ha −1 y −1 and with a CV of 0.20. The FRP LP , estimated as the difference between FRP Y and FRP G , was 2.17 ± 1.50 Mg ha −1 .
Considering that total carbon NPP for the Selva Piana experimental site in 2007-2008 was 11.01 MgC ha −1 y −1 [33], and that the amount of C allocated in fine roots was 3.92 MgC ha −1 y −1 (with C content of fine roots at 40.02 ± 1.92%), the contributions to total NPP by FRP Y , stem and branches, leaves, and coarse roots were 36%, 33%, 22%, and 9%, respectively.
The step-wise analysis results indicated LAI, related to basal area ( Figure S1), and SWC as the variables affecting FRP Y (FRP Y = −7.504 + 0.145 SWC + 2.342 LAI, R 2 = 0.928, p < 0.01). The commonality analysis suggested that 44% of the whole variability was affected by the pure effect of LAI, 20% related to SWC, and 36% was affected by the joint effect of the two predictors.

Spatial Variability of Soil Respiration
In the study period, SR was 1.49 ± 0.22 µmol CO 2 m −2 s −1 and ranged from 1.04 to 1.83 µmol CO 2 m −2 s −1 . We observed relevant variability at both spatial (among the 9 experimental plots) and seasonal (among the 11 SR campaigns) scales with mean CVs of 0.15 and 0.46, respectively.
In the study site, significant relationships between SR and FRP were observed. We found a highly significant effect of FRP on SR among the different plots considering both annual mean value (FRP Y, R 2 = 0.702, p < 0.01, Figure 3) and the datasets including growing (FRP G ) and leafless periods (FRP LP ) (R 2 = 0.842; p < 0.01; Figure 4). Conversely, among plots, we did not find any significant relationships between annual average SR and soil parameters reported in Table 1 (data not shown).

Fine Root Production and Its Contribution to NPP
FRP values of this study were within the range of 2.9 and 9.6 Mg ha −1 yr −1 reported for several beech stands in Europe [31,34]. Indeed, in an independent experiment carried out in the same period and site using isotope-labelled soil in-growth cores, a net annual root-derived carbon input to soil was estimated at 4 Mg C ha −1 yr −1 [33].
Moreover, our study suggests that not accounting for FRP LP could lead to an underestimation of fine root contribution to C-cycle that could be relevant, confirming previous findings [35]. FRP LP could be supported by the mobilization and use of carbohydrate reserves demonstrated in several studies [36][37][38]. In our experimental site, this hypothesis is corroborated by the decrease of starch and soluble sugars during winter [39,40].
Our results indicate that FRP Y is mainly dependent on LAI, which represents a proxy of ecosystem productivity and ground coverage [41,42]. This suggests that above-and belowground compartments are strongly connected at both local [43][44][45] and regional scales [46]. The positive relationships between FRP Y with both LAI and SR demonstrate the crucial role of FRP in connecting the forest structure and soil C fluxes. The role played by fine roots on soil C fluxes could be dependent on the forest development stages. Indeed, a finding similar to that shown in the present study was reported for an old mature beech forest [25], while no relationship was found between SR and fine root biomass in a young beech forest [47].
A previous study carried out in 1996 in the Selva Piana stand [31] estimated an FRP Y of 3.8 Mg ha −1 yr −1 , less than half of the current study, although the ingrowth cores method estimated lower values of FRP [48]. Furthermore, in the cited study, FRP Y contributed less to annual NPP (28% vs. 36%). These results might be only partially explained by the 37% increment of aboveground biomass and could be affected by the strong differences in precipitation regimes in the two sampling years, especially during the July-August period. FRP, as well its contribution to NPP, can vary depending on environmental factors [49][50][51], as suggested by the optimal partitioning theory where C allocation to roots can increase when plant growth is limited by water and/or nutrients [52,53]. However, our results suggested a double effect of water availability on FRP at different temporal and spatial scales. Water limitations could stimulate allocation to fine roots (i.e., 1996 vs. 2007), but at the stand scale, in case of water shortage, FRP could be positively stimulated by SWC. If so, this result confirms the observed positive relationship found between beech fine root growth and water availability driven by different precipitation regimes during drought years [51].

Spatial Variability of Soil Respiration and FRP
A large intra-site variability in SR rates was observed in several ecosystem types ranging from savanna, tropical, boreal, to temperate forests [11,18,47,[54][55][56][57]. In the present study, the spatial variability of SR was not related to the soil microclimatic environment, including T soil and SWC, confirming previous findings reported for three temperate European forests [58]. Hence, our results suggest that FRP plays a major role in determining the spatial variability of SR in a Mediterranean beech forest characterized by a closed canopy.
In addition, our data show a reduction of the FRP influence on SR (logarithmic regression) that could be related to the effect of soil water shortage on SR fluxes during the dry seasonal period, as previously demonstrated for the Selva Piana site [59].

Conclusions
This work described the spatial interactions among the forest structure and belowground C fluxes in a Mediterranean beech forest characterized by an almost complete canopy closure.
Fine roots played a relevant role in the ecosystem C-cycle, representing the main component of NPP (36%) and explaining about 70% of the annual soil CO 2 efflux variability inside the stand.
The results obtained in this study seem to indicate a functional mechanism to optimize source-sink C relationships in response to spatial variability of microclimatic drivers associated with changes of fine-scale forest structural traits. Forest structure and functionality are highly interactive; hence, an improved understanding of their relationships is fundamental to address forest adaptation and mitigation to climate change. Furthermore, as the structural features of the forest are derived from the past management, these results may inform adaptive forest management options.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4907/11/3/255/s1, Figure S1: Relationship between the basal area, representing the sum of the area (m 2 ) of the cross-section of stems measured at 1.30 m height, and Leaf Area Index (LAI) measured at the centre of each sampling plot, Figure S2: Relationship between annual fine root production (FRP Y ) and leaf area index (LAI). Each point is a sampling plot. Table S1: Forest structural parameters, soil parameters, and FRP in the 9 sampling plots (SP). Basal area (m 2 ) is the sum of the stem cross section areas of the n trees present in each plot; D max (cm) is the maximum diameter measured in the sampling plot tree height; LAI (m 2 m −2 ); T soil ( • C) is the average soil temperature measured during the SR campaigns (n = 11); SWC (%) is the annual average soil water content measured during the SR campaigns except February 2008 because of the snow cover (n = 10); SMN (%) is the organic nitrogen percentage in the mineral soil layer; SMC (%) is the organic carbon percentage in the mineral soil layer; SON (%) is the organic nitrogen percentage in the organic soil layer; SOC (%) is the organic carbon percentage in the organic soil layer; Litter amount (g DW m −2 ); Litter N (%) and Litter C (%) are the nitrogen and carbon percentage of litter, respectively; FRP Y , FRP G , and FRP LP (Mg Dw ha −1 ) are fine root production estimated at annual scale, during the vegetative season, and during leafless period, respectively; SR Y , SR G , and SR LP are the mean of the soil CO 2 effluxes measured during the whole study period, during the vegetative period, and during the leafless period, respectively.