Introduction

Globally, soil organic C (SOC) amounts to about 1,500 Pg C in the upper meter of soil, ranging from 3.0 kg C m−2 in arid climates to 80.0 kg C m−2 in organic soils of cold regions (Lal 2004), and most of this SOC is stored in forest soils (Dixon et al. 1994; Batjes 1996). Soils of areas with a Mediterranean climate are relatively poor in SOC (Rodríguez-Murillo 2001; Rodeghiero et al. 2011). This is possibly the result of dry periods and heavy rainfall events combined with the millenary and intensive land occupation, and the exploitation of the natural resources of the Mediterranean area (Ibáñez et al. 1995). The ongoing climate change could dramatically impact SOC stocks of areas with Mediterranean climate (Ojima et al. 1995; Davidson and Janssens 2006; Kirschbaum 2006; Knorr et al. 2005). In particular, if the expected increase in temperature, drought periods and heavy rainfall actually occur (IPCC 2007), it will likely result in further SOC losses especially in the most sensitive areas due to land degradation (Neilson and Marks 1994; Kimble et al. 1998). These possible SOC losses have detrimental effects on soil structure and soil fertility, contributing to global warming by further increasing the atmospheric concentration of carbon dioxide (CO2) (von Lützow and Kögel-Knabner 2009). In this context, forests represent a resource for contrasting hydrological problems and help in mitigating the greenhouse effect. In fact, by sequestering large amounts of atmospheric C, forests play an important role in the global C cycle and are thought to offer a mitigation strategy to reduce global warming (Schimel et al. 2001). Hence, characterization of the SOC stock in forest soils is of outmost importance to provide a basis for the assessment of the overall status of this pool in a land use type, which plays a relevant role in the global C cycle.

With more than 18 million hectares of forests, representing about 35% of the total country surface, Spain is the third country in terms of forest area within the European Union, contributing to about 12% of the total European forest areas (FAO-FRA 2010). Thus, assessing the SOC pool of this extensive territory is strongly needed. Nevertheless, estimating the SOC stock at national scale represents a challenge for most European countries. Indeed, assessments of SOC stocks at country scale are rather few because systematic sampling schemes including SOC, bulk density and stone content are quite rare (Morvan et al. 2008). Furthermore, a very high sampling density is required to get accurate estimates (Bellamy et al. 2005; Saby et al. 2008; Schrumpf et al. 2011).

The aim of this study is to estimate and compare the amount of SOC stored in the main forest biomes of Spain—conifer, broadleaf and evergreen broadleaf forests—in relation to the different climatic zones of the country. The definition of a SOC stock baseline is essential for future evaluations of the status of this pool. Furthermore, the identification of a baseline for the SOC stock in the different forest biomes can contribute to assess the variations in the SOC stocks occurring after land use changes from or to forest, i.e. changes in SOC stock due to afforestation/reforestation activities or conversion of forests into other land uses.

Materials and methods

The pedological database used in this study, denominated FOREDAF (Forestal-Edafologia), was created by the researchers of the Silviculture and Pasciculture Department of the Polytechnical University of Madrid, Spain. The data stored in the FOREDAF database refer to soil profiles that were described and sampled by pedogenic horizons in different surveys mainly across the 80’s and 90’s, and it has already been used for research purposes (Rubio and Escudero 2005). The FOREDAF database contains the analysis from 2,293 soil profiles and 7,558 pedological horizons. Each soil profile is characterized by its geographical coordinates, an accurate description of the vegetation, the parental material and soil chemical and physical properties (e.g. soil C concentration, pH, stone content and soil texture). The great value of this database is that soils have been collected with the same protocol both for sampling and analyses, resulting in a homogeneous database. To investigate the effect of climate on SOC stocks of the main forest biomes, the profiles were stratified according to the bioclimatic classification proposed by Rivas-Martínez (1983, 2005) and widely used under different climate types (Amigo and Ramirez 1998; Peinado et al. 1997; Mesquita and Sousa 2009). Bioclimatic belts, or zones, are the result of all physical variables affecting the landscape, since they are defined by thermal indexes and plant communities (Rivas-Martinez 1983). Moreover, these bioclimatic units correspond to an altitudinal zonation; therefore, bioclimatic belts might indicate differences in the SOC stock of forests, since they involve major factors such as altitude, slope and temperature patterns. Furthermore, this classification has been widely used for predicting the consequences of climate change on vegetation distribution (Sykes et al. 1996; Peng 2000; Walther et al. 2005). Figure 1 shows the seven bioclimatic belts of Spain along with the profiles distribution, while Table 1 reports the features of the seven bioclimatic belts based on temperature parameters. The 1,974 soil profiles considered in this study were assigned to the different bioclimatic belts using a climatic model developed for Spain (Sánchez-Palomares et al. 1999), which takes into account altitude, geographical position (x and y UTM Coordinates, Hayford ellipsoid) and the hydrographical basin or subbasin to which each profile belongs.

Fig. 1
figure 1

Map of the different bioclimatic belts of Spain (Rivas-Martinez 1983). Circles represent the location of the 1,974 soil profiles

Full size image
Table 1 Main features of the different bioclimatic belts of Spain based on temperature parameters
Full size table

Soil C stock calculation

The SOC stock was calculated for each horizon of the profiles according to Eq. 1 (Boone et al. 1999):

$$ y = a \cdot b \cdot c \cdot d $$
(1)

where y is the SOC stock per unit area (kg C m−2), a is the C concentration in the soil sample (kg C kg−1 soil), b is the soil bulk density (kg soil m−3), c is the depth of the horizon (cm) and d is the percent in mass of rock fragments [1- (% rock volume/100)].

All the SOC stocks were calculated for the mineral soil only, since information about the organic horizons were unfortunately missing. This obviously has to be considered since organic horizons contain a considerable amount of C. The SOC content was determined using the Walkley–Black procedure (Walkley and Black 1934). All the analyses were performed in the laboratories of the Silviculture and Pasciculture Department in Madrid. Because of the incomplete oxidation of carbon when using the Walkley–Black method, in this study, the SOC Walkley–Black values were converted to ISO standard, as already reported (Chiti et al. 2011; Ungaro et al. 2010), according to Eq. 2:

$$ {\text{SOC}}\_is{o_{14235}} = 0.0763 + 1.0288\,SOC{\__{W\& B}}\quad \left( {{R^2}\,{\text{of}}\,0.98} \right) $$
(2)

where SOCiso14235 is the SOC accordingly with the ISO standard and SOCW&B is the SOC as determined by the Walkley–Black method.

Since the soil bulk density, necessary to calculate the SOC stocks, was not present in the database, it was estimated using the pedotransfer function proposed by Adams (1973), as reported in Eq. 3:

$$ {\text{BD}} = \frac{1}{{\left( {a + b \times \% C} \right)}} $$
(3)

where BD is the bulk density, a and b are constants and%C is the soil C concentration. Many equations to estimate bulk density have been proposed for Spanish soils, from simple linear relationships (Saini 1966; Barahona and Santos 1981) to more complicated equations (Jeffrey 1970; Rawls 1983; Manrique and Jones 1991). Since most of the proposed equations have not improved significantly the correlation with BD (De Vos et al. 2005; Curtis and Post 1964), we have used the simple correlation reported by Eq. 3. To calibrate the function for the Spanish forest soils, a dataset of 592 samples having data of bulk density and C concentrations was created from different sources for the 0–30 and 30–100 cm depths of mineral soil. This dataset for bulk density includes profiles from some important soil databases such as WISE3 (Batjes 2008) and SPADE/M (Hiederer et al. 2006), which store soil profile data at global and European scale, respectively. The consistent over or under prediction of the bulk density estimated with Eq. 3 was evaluated by comparing measured and predicted values as reported by Moffat et al. (2007) and was taken into consideration when evaluating the error on the SOC estimates.

Rock fragment estimates were available for each horizon contained in the database. Taking into account the thickness of each horizon, the SOC stocks were referred to the 0–30 and 0–100 cm depths. When referring the stocks to the 0–100 cm depth, only the profiles reaching 1 m depth (1,951 profiles) were considered.

Total SOC in forests of Spain

The second Spanish National Forest Inventory (NFI2; DGCN 1998) was used to determine the contribution of each forest type to the total forest area of Spain. The NFI2 was chosen since data from the third national inventory (NFI3) are not yet available. Following the NFI2 definition, the areas with at least 5% of crown closure were considered as forest, except Dehesas with inter-spread crops. Canary and Balearic islands were excluded due to the absence of data in the database belonging to these areas. The mixed plots of the NFI2, representing around 10% of the total forest area of Spain, were assigned to one of the forest types considered in this study by taking into account the basal area of the tree species in the plot. Each of the sampling plots of the NFI2 (area = 1 km2) was assigned to a bioclimatic belt according to the climatic model of Sánchez-Palomares et al. (1999), and this allowed determining the total forest area of each combination of forest type and bioclimatic belt.

The total SOC stock (0–30 and 0–100 cm) stored in the Spanish forests was calculated by considering the surface occupied by the different forest types within each bioclimatic belt and their mean SOC stock as derived in this study. In some combinations of climate and forest type, the number of soil profiles was too small to obtain a representative mean value, particularly for Subalpine and Alpine conifer forests, Oromediterranean, Subalpine and Termomediterranean broadleaf forests, Coline, Montane and Oromediterranean evergreen broadleaf forests. The relative surface contribution of all these forest types is around 4% of the total forest area of Spain (DGCN 1998), and in the calculation, we used the mean value of each forest type.

Statistical analyses

The Kolmogorov–Smirnov test was used to verify the normality of the data, indicating that the hypothesis of normality can be rejected at 5% level, but not at 1% level. Outliers were excluded after the calculation of the median value. Significant differences (P < 0.05 and P < 0.001) between means of each combination of vegetation type and bioclimatic belt were tested by one-way analysis of variance and Tukey’s test (Sokal and Rohl 1997). To investigate the correlation between the stocks and other important parameters such as mean annual precipitation, maximum annual temperature and clay content, single regressions were performed for each of the forest biomes.

Results

Table 2 shows the fits of the pedotransfer function used in SOC stock calculation to a set of three functions for the 0–30, 30–100 and 0–100 cm depths. The minimum standard error (SE) of the estimates was calculated with Eq. 3: R 2 = 0.64 in the 0–30 cm depth and R 2 = 0.32 in the 30–100 cm depths. However, only considering the 0–100 cm depth, Eq. 3 fits more closely the data of the two parameters (R 2 = 0.66; P < 0.001) as also showed by Fig. 2. The distribution of residuals was much more uniform across C concentrations in this latter case than in the others, and therefore, Eq. 3 with the parameters obtained from the fit to the 0–100 cm depth was selected to estimate BD. The error in the estimate of the bulk density, evaluated by calculating the mean difference between measured and predicted values, suggests a small possible over- or underestimation of the real bulk density value of about 0.1 kg m−3.

Table 2 Results of some fits of bulk density versus C concentration for different soil compartments: 0–30, 30–100 and 0–100 cm
Full size table
Fig. 2
figure 2

Relationship between bulk density and organic C concentration for the 0–100 cm depth. Circles are the experimental values from 592 horizons, and the curve is given by the following equation: \( F = 1/\left( {a + b*C\% } \right) \). R 2 = 0.66 (P < 0.001)

Full size image

Table 3 shows the statistical characterization of the database for the three forest types considered in this study. Concerning the SOC stocks, the differences in the median values indicate a significant higher amount of C stored in broadleaf forests than under the other forest types, both considering the top 30 cm of mineral soil and the whole profile down to 1 m: 5.9 ± 0.1 kg C m−2 and 10.1 ± 0.3 kg C m−2, respectively. The SOC stocks of conifers and evergreen broadleaf forests were 5.6 ± 0.1 and 3.4 ± 0.2 kg C m−2 in the 0–30 cm depth and 10.0 ± 0.2 and 6.5 ± 0.4 kg C m−2 in the whole profile, respectively.

Table 3 Descriptive statistical analysis of the complete data set for the SOC stock of the 0–30 and 0–100 cm depths of mineral soil of the main forest types of Spain
Full size table

Forest SOC stocks in relation to different bioclimatic belts

Differences in the distribution of the SOC stocks were evident when the soil profiles were stratified according to the different bioclimatic belts. Conifers are the only forest biome represented in all bioclimatic belts. Considering the 0–30 cm depth and the whole profile, the differences in the median values among bioclimatic belts highlight statistically significant differences (P < 0.05; Fig. 3). Particularly, the average SOC stock of conifer decreases from the Coline (9.1 ± 3.5 kg C m−2) and Montane (6.8 ± 3.2 kg C m−2) belts to those zones with a more typical Mediterranean climate such as the Termomediterranean (3.0 ± 2.6 kg C m−2) and the Mesomediterranean (4.6 ± 2.7 kg C m−2) belts. The differences highlighted in the 0–30 cm soil depth were also found in the 0–100 cm depth (Fig. 3). Under broadleaf forests, the behaviour was similar to that of soils under conifer forests, the SOC stocks decreasing with increasing Mediterranean climatic conditions (Fig. 3). In the top 30 cm of soil, the forests in the Coline (7.6 ± 3.8 kg C m−2) and Montane (7.1 ± 2.8 kg C m−2) zones had higher stocks than the forests of the typical Mediterranean area such as the Mesomediterranean (4.6 ± 2.4 kg C m−2). The same trend was observed also in the 0–100 cm depth. Lastly, the evergreen broadleaf forests, widely present only in the Meso- and Supramediterranean bioclimatic belts, did not show significant differences in the SOC stocks both considering the top 30 cm or the whole profile. In the Mesomediterranean belt, the SOC stock was 3.5 ± 2.1 and 6.2 ± 4.3 kg C m−2 in the top 30 cm and in the whole profile, respectively, while in the same compartments of the Supramediterranean belt, the stocks were 3.3 ± 1.9 and 6.0 ± 4.2 kg C m−2, respectively (Fig. 3). Considering the whole profile, the SOC stock of the different forest biomes increases by about 50% in all bioclimatic belts, indicating that a high amount is stored also in the subsoil (30–100 cm depth). Within each bioclimatic belt, no differences were observed between conifer and broadleaf soils in term of C stored totally in the profile, while the evergreen broadleaf forests showed always a significant lower amount of SOC.

Fig. 3
figure 3

Soil organic C stock in the 0–30 and 0–100 cm depth of mineral soil under conifers (above), broadleaf and evergreen broadleaf forests (below) in the different bioclimatic belts: Azonal Riparian (Ar), Coline (C), Montane (M), Termomediterranean (Tm), Mesomediterranean (Mm), Supramediterranean (Sm), Oromediterranean (Om). Box indicates the mean (dashed line) and median (solid line), while the bottom and top of the box are always the 25th and 75th percentile. The ends of the box represent the minimum and maximum of all the data. Different letters indicate significant differences at P < 0.05. Numbers above the x axis represent the coefficient of variation (CV)

Full size image

The variability of the SOC stocks within the same forest biome was large and varied between bioclimatic belts (Fig. 3). In the 0–30 cm depth of conifer forest soils, the variability was lower with high C stocks (CV = 38% in the Coline and CV = 42% in the Montane zones) and increased when approaching the typical Mediterranean belts such as the Termomediterranean (CV = 68%). The variability increased only slightly when considering the whole profile, about 6% in the coldest belts and 4–13% in the Mediterranean zones. Under broadleaf forests, the variability was of the same order of magnitude as that found in soils under conifer forests, about 50% in the 0–30 cm depth and about 6–10% more when considering the whole profile. Evergreen broadleaf forests showed the highest CV in the 0–30 cm depth than the other forest types of the same belt, about 60%, and the CV increased by about 7–10% when considering the whole profile.

Within each forest type, a trend in the SOC distribution between bioclimatic belts was also supported by the significant positive correlation between SOC stocks and mean annual precipitation and the negative significant correlation between mean maximum annual temperature and SOC stocks (Fig. 4). At 0–30 cm depth, the SOC stocks showed the highest correlation with precipitations in soils under conifer forests (R 2 = 0.27; P < 0.001, Fig. 4) and the lowest in soils under evergreen broadleaf forests (R 2 = 0.05; P < 0.025). Opposite, the SOC stocks and the maximum temperature showed a negative significant correlation with an R 2 = 0.18 (P < 0.001) in soils under conifer forests and an R 2 = 0.02 (P < 0.097) in soils under evergreen broadleaf forests. Unexpectedly, the SOC stocks and the clay content of the three forest biomes showed a small but significant positive correlation (Fig. 4). The effect of clay was more evident in the evergreen broadleaf forests (R 2 = 0.24; P < 0.054) than in the conifer forests (R 2 = 0.10; P < 0.001) and broadleaf forests (R 2 = 0.10; P < 0.001), where the effect of clay was minimal. The same correlation performed considering the whole profile (0–100 cm depth) and the 30–100 cm depth did not improve the relationships (data not shown). Other parameters were related to the SOC stocks of the forest biomes, but all of them resulted to be poorly correlated. In particular, the correlation between SOC stocks and pH was negative in soils under conifer forests (R 2 = 0.01; P < 0.001) and evergreen broadleaf forests (R 2 = 0.04; P < 0.05). The correlation between the SOC stocks and the altitude was also poor with an R 2 of 0.02 (P < 0.05) for soils under broadleaf forests and an R 2 of 0.05 (P < 0.05) for soils under conifer forests.

Fig. 4
figure 4

Correlation between the SOC stocks of the three forest biomes and total annual precipitation (mm), maximum annual temperature (°C) and clay content (%) for the 0–30 cm depth of conifers (a), broadleaf (b) and evergreen broadleaf forests (c), respectively

Full size image

Total SOC in forests of Spain

Figure 5 shows the total SOC stored in the 0–30 and 30–100 cm depths of each forest biome for Spain. The highest total SOC stock (0–100 cm depth) was found under conifer forests 675.3 Tg of C, followed by broadleaf forests with 305.2 Tg of C and by evergreen broadleaf forests, 239.0 Tg of C. The amount of SOC was well distributed between the two depths. Particularly, the amount stored in the 0–30 cm depth varied from 55% under conifers to 58% under broadleaf forests. The contribution of subsoil (30–100 cm) C stocks was large, ranging from 45% under conifers to 42% under broadleaf forests.

Fig. 5
figure 5

Total SOC stock (Tg C) stored in the soil of the three forest biomes in Spain for the 0–30 and 30–100 cm depths. Different letters indicate significant differences at P < 0.05

Full size image

According to the stratification in bioclimatic belts and forest types, the highest amount of SOC to 1 m depth is found under conifers in the Supramediterranean (220 Tg C) and Mesomediterranean (205 Tg C) belts due to the large area conifer forests occupy in these regions. Broadleaf forests showed the highest total SOC amount in the Coline (78 Tg C), Montane (100 Tg C) and Supramediterranean (73 Tg C) belts, while the highest amount of SOC for evergreen broadleaf forests was observed in the Mesomediterranean belt, 144 Tg C.

Discussion

The differences between the mean SOC stocks of the three forest biomes in the bioclimatic belts suggest a high influence of climate features on the SOC stocks, overshadowing effects of soil texture. In the 0–30 cm depth of all forest biomes, clay content and SOC stocks had a small positive correlation. This is undoubtedly an unexpected result, which suggests a minimal influence of the soil texture on the capacity of soils to store C in Mediterranean areas. On the contrary, a major role of climate, particularly precipitation rather than the temperature, is suggested. Also, vegetation type appears to be an important factor. On the other hand, a strong correlation of the SOC stock with climate has been partly observed also in other studies. Rodríguez-Murillo (2001) found that annual precipitation was the best parameter correlating with SOC stocks in Spanish forests. The same correlation for forests of the Mediterranean area was also found by Rodeghiero et al. (2011), indicating a marginal relationship between SOC stocks and clay content. The higher mean SOC stocks are observed in the zones where the Atlantic conditions predominate such as the Coline and Montane belts, in agreement with the presence of the most productive forests of the country, composed mostly of pines and eucalyptus tree species (DGCN 1998). The rest of the country, where Mediterranean conditions predominate, shows the dominance of oak species in pure stands or mixed with pines and a wide variety of shrubs vegetation, and the average SOC stock decreases, showing its minimum in the Mesomediterranean belt under evergreen broadleaf forests.

Concerning the SOC stocks, our calculated values are well within those reported for Spanish forests. Very low values were reported by Diaz-Hernandez et al. (2003) for south-east forests of Spain with a semi-arid climate, 3.5 kg C m−2 to a depth of 50 cm. Balboa-Murias et al. (2006) in a mature broadleaf forest in north-west of Spain found a SOC stock of 6.4 kg C m−2 in the upper 30 cm of mineral soil. The results are also in good agreement with an estimate done by Rodríguez-Murillo (2001) for the 0–70 cm depth of conifers and broadleaf forests of Spain, 7.6 ± 6.7 and 9.4 ± 5.8 kg C m−2, respectively.

The comparison with SOC estimates of other countries of the Mediterranean basin indicates much higher values in the eastern Mediterranean countries. For Turkey, Evrendilek et al. (2006) reported an average SOC stock of 17.2 ± 2.6 kg m−2 to a depth of 30 cm for conifer forests, while in Greece, a SOC inventory conducted between 1979 and 1998 indicates an average SOC stock of forest soils for the 0–30 cm depth of 7.2 ± 3.9 kg m−2 (Rodeghiero et al. 2011). Lower stock is instead observed in southern France, where the SOC in the upper 30 cm of forest soils have been estimated, ranging from 1.9 to 5.6 kg C m−2 (Arrouays et al. 2001).

Given the high variation in the SOC stocks, about 95 profiles are required to estimate mean SOC stocks with a <10% error and a confidence level of 95%, assuming a normal distribution of the SOC stocks (Bernstein and Bernstein 1999). Under this assumption, SOC stocks for the considered forest types can be estimated rather accurately given the high number of profiles considered in this study. Particularly, the error in the SOC stock calculated in this study can be determined through the errors in factors which determine the SOC stock (i.e. BD,%C, stoniness). The relative error in C concentration is taken to be SE/(mean%C) = 3%. The relative error calculated in the same way for bulk density and rock fragments is 1% and 6%, respectively, thus leading to a total error in the SOC stock calculation of about 10%.

The value of SOC stored in the different forest types of each bioclimatic belt can allow a precise calculation of the total SOC stored in the forests of Spain. The total SOC stored in the upper 30 cm of the Spanish forests is 671 ± 10 Tg of C, while the amount stored in the soil down to 1 m is 1,218 ± 16 Tg of C (Fig. 4). Montero et al. (2005) quantified the total tree C for whole Spain as 720 Tg C. We have a similar amount in the top 30 cm of the mineral soil and about 70% more when considering the whole profile. In other words, of the ~2,000 Tg of C stored in forest systems, 60–65% is stored in the soil. Organic layers are not quantified. The fact that soil stores twice more organic C than the forest biomass is well known (Lal 2004), and this confirms the reliability of the estimate.

Conclusions

Both vegetation and climate play a major role in determining the SOC stock of Spanish forests, rather than the soil properties (clay content, pH) and other site parameters (altitude). Conifers and broadleaf forests store a sensibly higher amount of SOC than the evergreen broadleaf forests. The intensification of drought and extreme temperatures could cause an expansion of this forest type also where milder conditions now predominate, inducing possible SOC losses. Such SOC losses would be driven by both vegetation dynamics and change in climate conditions, being both closely interrelated.

Country-level soil databases provide extremely important information about current SOC baseline and can be used as a useful tool to evaluate future changes in SOC stocks under future climate conditions and/or land use change scenarios.