Soil properties and understory herbaceous biomass in forests of three species of Quercus Northeast Portugal

Aim of study: This paper aims to characterize some soil properties within the first 25 cm of the soil profile and the herbaceous biomass in Quercus forests, and the possible relationships between soil properties and understory standing biomass. Area of study: Three monoespecific Quercus forests (Q. suber L., Q. ilex subsp. rotundifolia Lam. and Q. pyrenaica Willd) in NE Portugal. Material and methods: During 1999 and 2000 soil properties (pH-KCl, total soil nitrogen (N), soil organic carbon (SOC), C/N ratio, available phosphorus (P), and available potassium (K)) and herbaceous biomass production of three forest types: Quercus suber L., Quercus ilex subsp. rotundifolia Lam. and Quercus pyrenaica Willd were studied. Main results: The results showed a different pattern of soil fertility (N, SOC, P, K) in Quercus forests in NE of Portugal. The C/N ratio and the herbaceous biomass confirmed this pattern. Research highlights: There is a pattern of Quercus sp. distribution that correlates with different soil characteristics by soil characteristics in NE Portugal. Q. pyrenaica ecosystems were found in more favoured areas (mesic conditions); Q. rotundifolia developed in nutrient-poor soils (oligotrophic conditions); and Q. suber were found in intermediate zones.


Introduction
The Iberian Peninsula is one of the main centres of diversity of the genus Quercus in Eurasia, with more than nine native species forming extensive forests and woodlands (Ruiz de la Torre, 2002). In the Northeast of Portugal, where the present study was carried out, forestland use occupies an approximate area of 156,000 ha (24% of the territory) (IFN, 2005). In this area, the most abundant species of the Quercus genus are Q. suber L., Q. pyrenaica Will and Q. ilex subsp. rotundifolia Lam occupying 12.5%, 11.5% and 9.4% of the forestland, respectively. Q. pyrenaica is a deciduous transitional Mediterranean oak, which is restricted to Southwest Europe (west-northwest Spain, southwest France and Northeast Portugal) and some isolated sites in northern Morocco.
Q. pyrenaica occurs where there is a transition between typical Mediterranean sclerophyllous and tempe-rate deciduous forests (Tarrega et al., 2006). It occurs in regions with an annual precipitation above 740-800 mm and a mean temperature around 12°C, on acid soils (Castro, 2008). The species is well adapted to survive summer water deficits while maintaining photosynthetic production (Rico et al., 1996). Therefore, it often occupies transitional areas from sub-humid to semi-arid conditions. Q. suber is an evergreen Mediterranean tree relatively abundant in the western parts of the Iberian Peninsula, as well as in western Italy, the large islands between Italy and Spain, and in North Africa (Morocco, Algeria and Tunisia) (Tutin, 1964). In the Iberian Peninsula, most Q. suber occur on acidic soils in the centre and western half of the Peninsula. Q. suber grows in non-carbonated soils such as acidic soils on granite, schist or sandy substrate and occasionally on neutral soils over dolomitic bedrocks (Pausas et al., 2006). It occurs in regions with an annual precipitation normally above 600 mm and a mean temperature around 15°C (Blanco et al., 1997). Quercus ilex is a sclerophyllous evergreen tree very abundant in a western Mediterranean region. It is occurs in the driest conditions, follo-wed by Q. suber. For instance, Q. suber dominates in coastal areas where the oceanic influence is stronger, while Q. ilex is characteristic of the driest areas on calcareous and even acid soils.
Quercus species are long-life trees and are frequently more than 100 years old (Plieninger et al., 2003). Over such an extended period, trees can significantly affect the fertility of the soil, mostly by recycling via leaf litter and root turnover of nutrients that were pumped from deep layers (Moreno et al., 2007). On the other hand, in the Mediterranean climate, nutrient availability is largely dependent on favourable combinations of warm temperatures, high soil moisture, and particularly of the quality and quantity of soil organic matter (Gallardo et al., 2000). Carbon accumulation in soil is strongly influenced by biological factors, such as vegetation (Vesterdal et al., 2008), microbial communities that decompose inputs (Ekschmitt et al., 2008), and by abiotic factors (Homann et al., 2007;Zinn et al., 2007). Nitrogen (N) is the most limiting nutrient in ecosystems (Whitehead, 1995) and the quantity of N in soils is intimately associated with organic matter levels, thus with soil quality. Assuming relatively low atmospheric inputs (EEA, 2011) and low losses via leaching or denitrification, soil N concentration basically depends on the balance between the N mineralisation rate and the N uptake by roots and soil microorganisms. Phosphorous (P) presents a mineral concentration that is quantitatively as important as the organic one, and forms part of soil minerals and rocks. Although P is required in lower amounts than other major nutrients (approximately 10 times less than for N and potassium) is another essential element for plants. P is unique among the anions in that it has low mobility and availability in soils. Potassium (K) is required by plants in approximately the same or slightly larger amounts as N. K does not form a part of any organic molecule, but plants usually amass large quantities of K in their tissues. K is mainly cycled through mineral weathering, but also by biological pumping through leaf litter decomposition (Lucas, 2001). Unlike N and P, K is easily leached from leaves and plant tissues (Aber & Melillo, 2001).
On the other hand, different processes such as weathering dissolution, atmospheric deposition, leaching, and biological mechanisms (immobilisation by plants and microbes) contribute to the vertical distribution of soil nutrients. It is known that leaching moves nutrients downward, while, biological mechanisms generally moves nutrients upwards. Previous studies showed that the most limiting nutrients for plants (N, P, K) were relatively more abundant in surface soil horizons than in deeper horizons (Gallardo, 2003;Jobbágy & Jackson, 2001).
The effect of tree species on forest soils is not fully understood and has been discussed for more than a century (Binkley, 1995). Also, tree species are often associated with specific soil types (Kristensen et al., 2004). For instance, at the landscape scale, it is known that broadleaved forests are mainly located on nutrientrich soils whereas coniferous forests are found in poor soils. Furthermore, several authors (cited by Zas & Alonso, 2002) argued that the relationships between vegetation and soil characteristics could be used as an important tool to assess sites. It has been recently reported that tree species is the most important explanatory variable for the C/N ratio in soils of European forests (Cools et al., 2014). Thus, it would be expected that the ecological functioning of the Quercus sp. studied could be expressed in different soil properties. In this study, we aimed to i) study soil properties and herbaceous biomass production on three different Quercus ecosystems located in the Northeast of Portugal; ii) study whether soil properties varied along depth in the three ecosystems; iii) associate these soil properties with herbaceous biomass production by each ecosystem.

Study area and experimental design
The study area was located in Trâs -os-Montes (NE Portugal), between the coordinates 41°23' N, 1°56' W and 41°50' N, 7°02' W ( Fig. 1). This region is characterized by high climatic and landscape diversity. Due to its internal situation, it has few maritime influence, yet its diversified geomorphology results in a wide variation of rainfall, this being the order of 1,500 mm in inner mountain areas in the north and, 400 mm in areas of the Upper Douro valley embedded. Among the native forests in the region, were chosen in this study, the three most representative species of Quercus: Quercus pyrenaica Will, Quercus ilex subsp. rotundifolia Lam (henceforth Q. rotundifolia) and Quercus suber L. Four experimental sites in Q. pyrenaica and Q. suber and three sites in Q. rotundifolia forest were selected (Fig. 1). Each site was representative of each species forest type and was selected 426 M. Castro and E. Fernandez-Nuñez / Forest Systems (2014)  randomly across the distribution area. A total of 68 plots were sampled: 44 plots in Q. pyrenaica (11 plots per site); 12 plots in Q. rotundifolia (4 plots per site); and 12 plots in Q. suber (4 plots per site). Selected forests showed physiographic and ecological dendrometer measurements that were very different from each other ( Table 1). The study area is characterized by a Mediterranean climate with an annual rainfall between 520-1386 mm and annual average temperature between 11.9-14.2°C (INMG, 1991). The average altitude of the study areas was between 380-980 m a.s.l. Density varied considerably among the different ecosystems; ecosystems of Q. rotundifolia had higher plant densities, while Q. suber ecosystems were more dispersed (Teixeira et al., 2001). During 1999 and 2000 we proceeded to study the characteristics of the soil and herbaceous biomass of three forest types: Q. suber, Q. rotundifolia and Q. pyrenaica.

Sample collection and analysis
A total of 204 soil samples were randomly collected using a soil corer: 132 soil samples were collected un-der Q. pyrenaica (11 samples at each sites × 4 sites × 3 depths), 36 samples under Q. suber (4 samples at each site × 3 sites × 3 depths) and 36 samples under Q. rotundifolia (4 samples at each site × 3 sites × 3 depths). Litter was removed from the surface before sampling. Upon collection, samples were taken to the laboratory to be dried to constant weight (40°C) and sieved through a 2-mm sieve. The < 2-mm fraction was used to determine the main soil properties: pH-KCl was determined by the potentiometric method (Santos, 1965); Total soil nitrogen (N) by the Kjeldahl method (Bremner, 1996), and soil organic carbon (SOC) by the Walkley-Black method described by Schulte (1980). Available P and K (mg kg -1 ) were extracted with an ammonium acetatelactate A-L solution (pH 3.7; ratio 1:20). Available P and K were determined by spectrophotometry and flame photometry, respectively by the Egner-Riehm-Domingo method (Egner-Riehm and Domingo, 1960).
Herbaceous biomass was evaluated by the method of cutting and weighing in square 0.25 m 2 . A total of 68 samples were cut: 44 samples in Q. pyrenaica (11 samples at each site × 4 sites), 12 samples in Q. suber (4 samples at each site × 3 sites), and 12 samples in Q. rotundifolia (4 samples at each site × 3 sites) in the late spring of 2,000. Herbaceous biomass was cut at a height of  2.5 cm. Subsequently, each sample was dried (48 at 65°C) and we proceeded to determine the dry matter yields.
In the winter of 1999, a different number of trees (n) were measurement in each plot (Table 1). All the trees in each sample plot were labelled with a number. Cross diameters at breast height (D: 1.3 m above ground level) were measured with callipers, and the arithmetic mean was calculated. The following stand variables were calculated for each plot: number of trees per hectare (N); average diameter (D); and stand basal area (BA).

Statistical analysis
The variables studied were: pH-KCl, total soil nitrogen (N); soil organic carbon (SOC); available P; avai-lable K; C/N ratio; and herbaceous biomass. For soil variables, data were analyzed via Repeated Measures ANOVA, using Mauchly's Criterion to test sphericity. If the sphericity assumption was met, univariate approach output was used; if not, multivariate output (Wilks' Lambda test) was used. The LSD (Least Significant Difference) test was used to make subsequent pairwise comparisons after a normalization test (α = 0.05), but only when factors were significant according to the results of the ANOVA analyses. The statistical software package SAS (2001) was used for all analyses. A Pearson correlation test was used to evaluate possible correlations between soil variables and biomass.
Principal component analysis (PCA) with the topsoil variables measured and biomass was carried out in the CANOCO 4.5 program (ter Braak & Smilauer,428 M. Castro and E. Fernandez-Nuñez / Forest Systems (2014)  2002). To improve the interpretability of principal components we used the orthogonal varimax rotation that results in uncorrelated principal components. Results of the multivariate analysis were visualized in the form of a bi-plot ordination diagram created with CanoDraw © software.

Results
The herbaceous biomass production obtained was significantly lower (F: 14.69, dg: 2, p < 0.001) in the ecosystems of Q. rotundifolia compared with Q. suber and Q. pyrenaica ecosystems (Fig. 2).
The interaction ecosystem × depth was significant for pH-KCl, total soil N, available P, available K and SOC (Table 2). Q. pyrenaica and Q. suber ecosystems showed lower pH-KCl values than Q. rotundifolia ecosystems. The difference between Q. rotundifolia and the other two ecosystems studied was about 0.5 pH-KCl units (Fig. 3) Soil pH-KCl was significantly reduced at 15-25 and 5-15 cm compared with the top 5 cm soil layer in Q. rotundifolia and Q. pyrenaica ecosystems, respectively (Fig. 3). No differences between depths were found under Q. suber ecosystems. On the other hand, concentration of macronutrients and SOC consistently decreased with depth in the three studied ecosystems (Fig. 4). In the three ecosystems, more than 50% of SOC was stored in the topsoil while, the lower amounts were found in the 5-15 cm (26%) and 15-25 cm layer (20%). When analysing the influence of ecosystems, the highest SOC values were observed under Q. pyrenaica at 0-15 cm and under Q. pyre-naica and Q. rotundifolia in the 15-25 cm layer. In general, Q. suber ecosystems showed lower SOC values along the soil prof ile compared with the other two ecosystems studied, with the exception of the top 5 cm layer where no differences between Q. suber and Q. rotundifolia ecosystems were found.
Q. rotundifolia ecosystems showed higher C/N ratio at 15-25 cm compared with the top 5 cm soil layer (Fig. 5), while the results showed that in Q. pyrenaica ecosystems, the C/N ratio was significantly increased at 5-15 cm compared with 15-25 and the top 5 cm soil layer. No effects of depth on the C/N ratio were found in Q. suber ecosystems. On the other hand, Q. rotundifolia showed a higher C/N ratio than Q. suber and Q. pyrenaica ecosystems along the soil profile.
The interaction ecosystem x site was significant for pH-KCl, total soil N, available P, available K, SOC and C/N ( Table 2). The results showed that pH-KCL was significant higher under Q. rotundifolia sites (Lagoa, Sendim and Vila Boa) compared with Q. pyrenaica and Q. suber sites (Table 3). In general, the higher total soil N and available K levels were found in Q. pyrenaica sites along the soil profile. Rebordainhos was the site with the higher available P at the top 5 cm compared with Q. suber and Q. rotundifolia sites (except Lagoa). On the other hand, SOC was significant higher in the four Q. pyrenaica sites studied compared with Q. suber and Q. rotundifolia sites at the top 5 cm. Vila Boa (under Q. rotundifolia) showed the higher C/N ratio at 0-5 and 15-25 cm compared with the other sites studied while, Zido (under Q. pyrenaica) was the site with the higher C/N ratio at 15-25 cm.   Table 4 shows the results of the Pearson correlation test used to evaluate possible correlations between soil variables and biomass in the surface layer (5 cm). The highest correlations were found between N and SOC (r = 0.909), and N and P (r = 0.710), followed by SOC and P (r = 0.634). On the other hand, available K was positively correlated with N (r = 0.554), SOC (r = 0.588), and P (r = 0.448). Moreover, a moderate negative correlation was found between N and C/N (r = -0.447) and herbaceous biomass and C/N ratio (r = -0.367). As with the surface layer, a positive correlation between N and SOC (r = 0.847), K and N (r = 0.659), K and P (r = 0.659), K and SOC (r = 0.503), and N and P (r = 0.515) was found in the deeper layer (Table 5).
PCA was significant (p < 0.000) in the explanation of dependent variables (Fig. 6). PCA showed that 92% of the total variance of the data could be explained using only two principal components. The first axis,  PC1 was positively related to total soil N, SOC, P, K and negatively related to C/N ratio, indicating the existence of a fertility gradient. On the other hand, the positive contributions of herbaceous biomass and the negative contributions of the C/N ratio suggested the existence of a gradient of productivity associated with soil C/N ratio. Moreover, PC1 separated N, SOC, P, K, which were concentrated on the positive side of axis 1, reflecting that these parameters reached higher values in the case of Q. pyrenaica ecosystems and lower values in the case of Q. rotundifolia ecosystems. The second axis, PC2, was positively correlated with biomass and negatively with the C/N ratio.

Discussion
It is known that Mediterranean soils often suffer from nutrient def iciencies, especially of N and P (Sardans et al., 2004). In our study, the soils of the three ecosystems are defined as acidic, which could affect the availability of N, P, K and the decomposition rate of the organic matter, and therefore the accumulation of SOC (Ulrich & Summer, 1991). In the three ecosystems studied, nutrient scarcity manifested as low N content along soil profile. With the exception of the top 5-cm layer of Q. pyrenaica ecosystems, all soil layers showed amounts below 0.27 mg g -1 , far lower than the worldwide mean N value for the first 30 cm of soil proposed by Batjes (1996) that was settled to 4.2 mg g -1 for Leptosols, or 4.9 mg g -1 for Litosols and humid Cambisols in Spain (Zas & Alonso, 2002).  It is known that N contents reaches higher values in soils developed over schist than over granites (Turrión et al., 2009). This could explain the fact that the total soil N found in this study was higher than the levels reported by Silla & Escudero (2006) and Turrión et al. (2009) in Q. pyrenaica forests developed over granites. In the case of Q. rotundifolia and Q. suber, total soil N contents were between the range obtained by Canadell and Vilá (1992) for Q. ilex forests studied in Spain & Otieno et al. (2011) for Q. suber forests in Southern Portugal. Furthermore, our results showed similar available P contents than those reported by Rapp et al. (1999) in Q. pyrenaica forests developed over schist; and higher levels compared with those obtained by Andivia et al. (2010) for Q. suber forests developed over a sandy soil with low cation exchange capacity which could have favoured cation leaching. Regarding K, the results showed higher K than P levels in the three ecosystems and along the soil profile. This could be explained because, in all ranges of soil pH, phosphate ions tend to react with other cations to form less soluble compounds; therefore, P availabi-432 M. Castro and E. Fernandez-Nuñez / Forest Systems (2014) 23 (3): 425-437 Table 3. pH-KCl, total soil N (mg g -1 ), available P (mg kg -1 ), available K (mg kg -1 ), soil organic carbon (SOC) (g kg -1 ), and C/N ratio in the each site studied (Q. pyrenaica-4 site, Q. suber-3 sites and Q. rotundifolia-3 sites) and in the three soil depths studied (0-5, 5-15 and 15-25 cm) lity could be reduce (Turrión et al., 2008). However, this reactions can be very pronounced in acidic soil because, in this case, Al is the dominant ion that will react with phosphate, as well as some Fe and Ca phosphates, which will result in very insoluble compounds of phosphate that are generally not available to plants.
On the other hand, unlike P, the enhancement of K availability in soils is easily achieved, because this nutrient remains stored in soil cation exchange sites (Bortoluzzi et al., 2005). Generally, K adsorption in soils is sufficiently strong to avoid the leaching process and sufficiently weak to supply this nutrient to the soil solution.
Soil, biomass in Quercus species 433 Table 5. Pearson correlation coefficient matrix among each pair of variables considered in the study for deeper layer (5-25 cm) (n = 136) N SOC pH-KCl C/N P K (mg g -1 ) (g kg -1 ) (mg kg -1 ) (mg kg -1 ) N (mg g -1 ) 1.000 SOC (g kg -1 ) 0.847*** 1.000 pH-KCl -0.103 -0.190 1.000 C/N -0.408*** -0.028 0.148 1.000 P (mg kg -1 ) 0.514*** 0.358*** 0.153 -0.149 1.000 K (mg kg -1 ) 0.659*** 0.503*** -0.018 -0.355*** 0.633*** 1.000 *** Significance at p < 0.001. Regarding to vertical distribution of macronutrients for plants (N, P and K), concentrations were higher in the top 5-cm soil layer and decreased down the soil profile in the three ecosystems studied. These results support the prediction that the most limiting nutrients for plants would have the shallowest distributions (Jobbágy & Jackson, 2001). In the case of vertical distribution of SOC content, it is known that soil-buffering capacity may reduce the effects of environmental variables (e.g. temperature and moisture) on SOC content in deep soil layers. Authors such as Jobbágy & Jackson (2000), who analysed more than 2,700 worldwide distributed soil profiles, found evidence suggesting that the vertical distribution of SOC is conditioned by the root distribution along the profiles. It is known that litter, dead materials, and biological activity accumulate at the soil surface, leading to the formation of soil organic matter (Melillo et al., 1989), and hence to organic carbon contents in the top soil layer (Jobbágy & Jackson, 2000).
On the other hand, the results obtained showed the existence of a fertility gradient among the three ecosystems: Q. pyrenaica > Q.suber > Q. rotundifolia. Thus, Q. rotundifolia ecosystems would be found in nutrient poor soils, in other words, in less favoured zones. Cools et al. (2013) identified tree species as the major factor explaining C/N ratios in European forest. Also, Quercus sp. showed an ample range of C/N ratio distributions; the deciduous oak had low C/N ratios while evergreen oak (Q. rotundifolia and Q. suber) had high C/N ratios. The authors also found that forest of Q. pyrenaica, oak specie with intermediate characteristics between the Eurosiberian and the more sclerophyllous Mediterranean species, showed intermediate C/N ratios. Our results agree with these authors for Q. pyrenaica and Q. rotundifolia. However, in the case of Q. suber, our results are not consistent with the mentioned authors, which could be explained by the different scales of these studies (European vs. Northeast of Portugal). On the other hand, it is known that plant species differ in their ability to modify soil properties (Vinton & Burke, 1995, Cornelissen et al., 1999. These changes may be due, besides climate and edaphic properties, to the growth form, life cycle, and chemical properties that can significantly affect the decomposition of soil organic matter and nutrient dynamics (Hooper and Vitousek, 1998;Carrera et al., 2009). The high C/N ratio found in Q. rotundifolia ecosystems (> 15) would result in a significant reduction of biological activity and, therefore lower rates of mineralization. This situa-tion, and the fact that Q. rotundifolia ecosystems present higher canopy density (greater constraints of light) and lower soil nutrient content than Q. suber ecosystems could explain the low herbaceous biomass production found under Q. rotundifolia. The higher herbaceous biomass found in Q.pyrenaica and Q.suber ecosystems (almost twice that found under Q. rotundifolia) also contributes positively to higher soil nutrient contents. It is known that plants with rapid growth such as grasses show a faster rate of decomposition, due their high concentrations of N and low concentrations of secondary compounds (Carrera et al., 2009).
It is known that slow-growing species as evergreen sclerophyllous can incorporate significant quantities of nutrients and soil organic matter in the soil (Gallardo, 2003;Moreno et al., 2007), as well as a high concentration of secondary components (Aerts and Chapin, 2000;Satti et al., 2003). In these situations, decomposition rate is very slow reflecting oligotrophic conditions in the case of Q. rotundifolia ecosystems.
Unlike the P and K, N dynamics in soil is not regulated by a chemical equilibrium, but primarily by biological processes, derived from soil biological activity. Thus, growth, productivity, and nutrient balance can mainly depend on the composition and soil microbial activity (Salazar et al., 2011), which in turn is influenced by weather conditions directly through moisture and soil temperature (Ley et al., 2004;Nielsen et al., 2009). Aponte et al. (2012Aponte et al. ( , 2010 compared soil development under Quercus. canariensis (marcescent) with that under Q. suber. They found higher N and SOC in soils developed under marcescent species, such as Q. pyrenaica in our study Furthermore, the higher nutrient concentration in Quercus sp. is present in the leaves (Salazar et al., 2011). The amount of leaf nutrients is of great importance in the case of Q. pyrenaica (marcescent) because the leaves are subject to annual shedding cycles, through which leaf biomass, and therefore nutrients, returns to the soil (Gallego et al., 1993) and therefore nutrients. Moreover, it is known that litter under Q. ilex and Q. suber has a very slow decomposition rate, resulting in thick, complex, and structured humus, which increases the soil binding capability for nutrients (Alarcón-Gutiérrez et al, 2009)

Conclusions
This study suggests that in Northeast of Portugal there is a pattern of Quercus sp. distribution that 434 M. Castro and E. Fernandez-Nuñez / Forest Systems (2014) 23 (3): [425][426][427][428][429][430][431][432][433][434][435][436][437] correlates with different soil characteristics by soil characteristics. Q. pyrenaica ecosystems were found in more favoured areas (mesic conditions); Q. rotundifolia developed in nutrient-poor soils (oligotrophic conditions); and Q. suber were found in intermediate zones. Future research should be designed to test other factors (climate, forest structure, and management regimens) in a more extensive area that clearly stated the existence of Quercus sp. distribution pattern in a relation to soil fertility.