Evaluating carbon stocks in soils of fragmented Brazilian Atlantic Forests (BAF) based on soil features and different methodologies

Brazil’s Atlantic Forest (BAF) is a highly fragmented, strategic environmental and socio-economic region that represents the fourth biodiversity hotspot while also producing many commodities that are exported globally. Human disturbance plays a pivotal role as a driver of BAF’s soil dynamics and behaviors. The soils under Late Primary and Secondary Semideciduous Seasonal Forests (LPSF and LSSF) were characterized by high to moderate resilience, with improved chemical properties as human disturbance decreased. The Transitional Forest to Cerrado (TFC) had the worst soil conditions. Disturbed Primary and Secondary Semideciduous Seasonal Forests (DPSF and DSSF) represent a transitional stage between LPSF/LSSF and TFC. Accordingly, SOCs stocks increased from TFC << DPSF, DSSF < LPSF, LSSF. In BAF soils, to avoid unreliable data, SOCs measurements should be (i) conducted to at least 1 m soil depth and (ii) quantified with a CHN analyzer. Human disturbance strongly affected the positive feedback between vegetation succession, SOCs, and soil nutrition. Soil development decreased as human disturbance increased, thus negatively affecting SOCs. Soils in the BAF require a long time to recover after the end of human disturbance, thus suggesting that preservation strategies should be prioritized in remnant BAF fragments.

Overall, taking into account all the previous considerations, this research aimed to: (i) quantify C stocks in BAF soils by (ii) comparing WBm vs CHNa methods, along (iii) different soil pedosystems (investigated to 1 m soil depth) featuring (iv) five different BAF fragments (in terms of human-disturbance magnitude) belonging to one of the most important, human-affected, and fragile ecosystem worldwide.A full suite of soil physical and chemical analyses was conducted in addition to SOC quantification to understand (vi) how soil features and different methodologies influence SOCs assessment.The present paper first combined and focused on all these related and pivotal aspects in one single research.

Soil organic carbon stocks (SOCs)
Table 3 reports the SOCs values, for each of the five investigated BAF fragments, also comparing the used methodologies for SOC quantification.
The largest difference in SOCs amounts was observed for DPSF (Degraded Primary Semideciduous Seasonal Forest) with values of SOC extracted by CHNa 99% higher than SOC values obtained with the WBm.The smallest differences were observed for both investigated Late Forests (LPSF: 14%, LSSF: 19%).Soil organic carbon stocks and CO 2 e, when quantified through the CHNa, were 40% higher on average, considering all investigated physiognomies to 1 m soil depth, when compared with the WBm.
As previously observed (Table 3), the WBm showed a mean value of 128.6 ± 14.3 t ha −1 of SOCs stocked up to 1 m soil depth.Of this, 62% (79.7 t ha −1 ) was concentrated in the first 0-40 cm, while the remaining 38% (48.9 t ha −1 ) was in the 40-100 cm soil depth.By using CHNa extraction, SOCs mean values up to 1 m depth were 180.9 ± 18.6 t ha −1 (Table 3).In this case, 58% (104.9 t ha −1 ) was in the first 0-40 cm soil depth, while the remaining 76.0 t ha −1 (42%) was found in the last 40-100 cm soil depth.Table 4 shows SOC values according to the method of SOC measurement, BAF fragment, and investigated soil depth.Looking at the SOC content vs soil depth, considering all investigated BAF fragments, an inverse correlation was observed, with very similar behaviors for both applied SOC extraction methodologies, even if the WBm resulted in a lower estimate of SOCs compared to CHNa.Also in this case, the LPSF and LSSF fragments had the highest (p < 0.05) SOC contents, regardless of the investigated depth.The DPSF, DSSF, and TFC do not show statistically significant differences except when SOC was analyzed through CHNa.Indeed, in this case, at least for the soil depth to 60 cm, a clear statistical difference was observed, with the following trend in SOC contents: LPSF ≥ LSSF >> DPSF ≥ DSSF >> T FC.Soil organic carbon stocks at the 80-100 cm depth were three (CHNa) to two (WBm) times lower compared to the surface A horizons (0-20 cm).

Factor analyses
The factor analyses (FA, Table 5) extracted a three-component model of robust statistical performance (all eigenvalues are over 1).The three extracted Factors explain 80% of the overall variance, meaning high statistical reliability.The first factor (F1) explained 56% of all observed variance in the data.F1 was positively concordant with pH, SOM, C-CHN, N, P, all basic exchangeable cations (K + , Ca 2+ , Mg 2+ ), CEC, BS, Cu, Mn, and both SOCs (regardless of the used method), while Al 3+ was negatively correlated.F2 (14%) was positively correlated with SOM, C-CHN, Fe, and both SOCs.Finally, F3 (10%) showed BD as inversely correlated to C-CHN, N, K+, S, B, and both SOCs.

Principal component analyses
Figure 1 shows the biplot from PCA.Overall, the two main components explain alone the 69% of all observed variability.All investigated soil parameters (arrows) are over 1, thus showing a high statistical significance inside the model.The five mains investigated BAF's physiognomies (points rounded by areas underlined by different colors) are grouped as: LPSF, LSSF ≠ DPSF, DSSF ≠ TFC.Table 1.Soil selected physical-chemical properties in the investigated BAF's fragments (mean ± SE).Different lowercase letters along columns (inside the same BAF fragments) showed significant statistical differences for different soil depths at p < 0.05.Different capital letters along columns (among the five different BAF fragments) showed significant statistical differences for the same soil horizons at p < 0.05).

Soil physical-chemical features
Features of investigated soils among the five different BAF fragments show interesting similarities and differences.
In most of the analyzed soil physical-chemical properties, we observed that LPSF, LSSF ≠ DPSF, DSSF ≠ TFC.Regarding human effects, past disturbances played a pivotal role in creating such clear differences.As argued by Roder et al. 27 , in more natural conditions like the LPSF and LSSF in this study, soils develop under mature forest vegetation, with higher spatial canopy cover and a well-developed organic surface horizon (O).Thus, O horizons can release basic cations in greater amounts, increasing soil pH 44 , while the forest canopy intercepts precipitation and limits leaching processes, thus limiting the loss of more soluble nutrients (basic cations) 45 .
Results from the investigated soil physical-chemical parameters show the pivotal role of human disturbance as a driving-factor in affecting pedoenvironment dynamics and behaviors.Even if pedoenvironments like those featuring LPSF and LSSF and somewhat even DPSF and DSSF are characterized by high to moderate resilience, with most of the investigated parameters improving as human disturbance decreased, recovery seems to be highly time-dependent.Indeed, the best soil conditions were found in those BAF fragments with no history of deforestation, exploitation for livestock, or agriculture, at least during the last century (LPSF), while the worst in the TFC, where humans acted as driving force until recently (30-40 years ago).Such observations suggest an Table 3. Comparisons in SOC stocks (SOCs) according to investigated BAF's fragments and used method for SOC stock quantification (CHNa CHN analyzer, WBm Walkley-Black method).Legend as in Table 1.Different letters, along rows and columns, showed significant statistical differences (p < 0.05).

Soil organic carbon stocks (SOCs)
The carbon concentrations measured using CHNa were significantly higher than those measured using WBm, with mean values of 180.9 ± 18.6 t ha −1 and 128.6 ± 14.3 t ha −1 respectively.On average, CHNa measurements of SOC were 40% higher than WBm, with the greatest difference in the DPSF plots, where SOC was 99% higher with CHNa.Such results demonstrate that CHNa can play a pivotal role in detecting significant differences in SOCs values among different environments.The Walkley-Black method is less able to quantify SOC, and thus may under-or over-estimate SOC change when comparing between disturbed environments or between different management strategies.This outcome is in line with studies conducted worldwide that reported an underestimation in SOC values by using WBm compared to CHNa 21,[48][49][50][51] .This is an intrinsic problem of the WBm, since it exhibits great variability in the dichromate oxidation degree and efficiency depending on factors such as SOC composition, soil bio-chemical-physical features, land cover and use, investigated soil horizon, etc. [47][48][49][50][51][52] .The CHNa detected higher SOC concentrations than the WBm, especially in soil horizons of BAF fragments with higher C content (Tables 1, 2).Similar results between the two methods were observed in lower C content conditions, as for the TFC or for deeper soil horizons.The decreased efficiency of the WBm compared to CHNa can be explained by the well-known affinity between iron-enriched tropical soils for OC 53 .Thus, CHNa was a more reliable method than WBm, particularly in soil horizons with larger SOCs due to mineral adsorption processes.Another factor could be associated with past land uses.Indeed, Davis et al. 54 reviewed the SOC methods for tropical soils of Brazil, showing that in more anthropized or less natural areas (like those featured by relatively recent fires) WBm resulted in smaller and more variable results, especially for soil surface mineral A and Ap horizons.Thus, the authors concluded that underestimation of SOC content in such soils should be expected with WBm measurements.Segnini et al. 55 and Tivet et al. 56 confirmed that in the tropical soil of Brazil, WBm strongly underestimated SOC contents; Tivet et al. 56 also identified historical land use as a factor affecting the efficiency of WBm.Due to the concomitant combination of several issues (vide supra), conversion factors are extremely sitespecific 56 , thus demonstrating the need to develop models calibrated for each investigated pedosystem to achieve a reliable comparison among SOC stocks.Despite SOC underestimation, WBm was the only method available to researchers in many cases; consequently, developing site-specific correction factors represents a strategy to improve data quality.Researchers worldwide proposed the application of a large range in correction factors (from 0.09 to 2.21) to increase WBm reliability 51 .A contradiction arises from the literature that authors often report the need to use CHNa to calibrate WBm measurements; this was done by sending samples outside investigated countries with related costs 22 .So, even if it is still widely used as a relatively low-cost method, it frequently requires a review/calibration process due to the large variability in the conversion factor.For instance, Gessesse and Khamzina 22 revealed that in soil samples collected in several locations in Ethiopia, the most common correction factors did not improve the reliability of WB vs CHNa derived results.Thus, they proposed using the Bland and Altman method 57 , commonly applied in clinical research, for assessing the conversion factor, concluding that a correction factor of 1.32 for non-calcareous, carbon-poor Ethiopian soils can be considered reliable.This research, as with many others [58][59][60][61][62][63][64][65][66] (vide supra), confirmed that correction factors are extremely site-specific and, thus not easily applicable without a double-check control with more expensive and sophisticated laboratory tools.As concluded by the review of Pribyl 67 , any factor used for OC conversion in SOM cannot be assumed as a universal constant; indeed, such numbers may be influenced by a combination of many factors (vide supra) that have the potential for serious error.Consequently, even if in the absence of other alternatives WBm often remains the only possibility to assess SOC, it is urgent to find new solutions that increase the reliability of data in unfavored socio-economic conditions.From this point of view, participatory and networking-based research promoted by Universities, Research Centers, Governments, and other institutional bodies such as non-governmental organizations (NGOs), represent opportunities to remedy this gap in data reliability.Implementing such kinds of projects is not only of scientific importance but also a moral question 22 .Indeed, GHG emissions, prevalently produced by developed countries, are responsible for climate change, mainly affecting developing ones 68 .
Looking at the SOCs vs soil depth, the LPSF and LSSF forests showed the highest values in all investigated depths compared to the other BAF fragments.In particular, if compared to the TFC (the most degraded fragment or the least natural one), LPSF and LSSF stocked around 4.5-5 times more SOC across each depth for both CHNa and WBm methods.The DPSF and DSSF seem to represent an intermediate stage between the aforementioned physiognomies.Depending on the measurement method (CHNa or WBm), DPSF and DSSF contained 2.5-3.5 times less SOC than LPSF and LSSF, but contained 1-1.5 times more than TFC.
Our results also show the deficiency in sampling and investigating the first 0-30 cm to quantify SOC 5 , which is the IPCC suggested minimum sampling depth for SOCs quantification.As argued by Jandl et al. 9 , considering only surface soil can dramatically underestimate overall SOC quantification, and sampling activities to at least 1 m soil depth can reveal additional important information.The present research confirmed that: (i) the SOCs between 40 and 100 cm were only slightly lower (≤ 10%) than stocks found in the first 0-40 cm; (ii) by extrapolating SOCs data from 30 to 100 vs 0 to 30 cm soil depth, the amount is significantly comparable, confirming Muños-Rojas et al. 69 suggestion of soil sampling up to 1 m depth for SOCs quantification.Indeed, if SOC stock www.nature.com/scientificreports/into the 30-100 cm soil depth was not measured, up to 50% of overall SOC (along the soil profile) would not be quantified.We observed a clear increasing trend with decreased human disturbance level in both SOC stocks and CO 2 e amounts.As argued by Chenu et al. 70 , under most preserved environmental conditions, SOC stock can approach a long-term equilibrium, with SOM inputs coming from vegetation and animal residues, roots and their exudates, etc., balanced by its degradation from mineralization processes.This equilibrium can be shifted, as observed in the investigated areas, by changes in land management.Our results agree with Roder et al. 27 , which demonstrated a decrease in SOM as human disturbance increased in fragmented BAF.Souza et al. 47,71 demonstrated that BAF recovery resulted from a large increase in SOCs after stopping human disturbance.In the present study, we demonstrated that such an increase in SOM was particularly found in Late Primary Semideciduous Forest (LPSF) and Late Secondary Semideciduous Forest (LSSF), which contained a significantly higher amount of SOCs compared to Disturbed Primary Semideciduous Forest (DPSF), Disturbed Primary Semideciduous Forest (DSSF), and the Transitional Forest to Cerrado (TFC).Previous research by Britez 72 confirmed that among BAF physiognomies, the Disturbed Semideciduous Seasonal Forest can be characterized by low SOCs stock (between 31.7 and 37.5 t ha −1 up to 100 cm soil depth), especially in the case of relatively recent human disturbance.The results from our present research validate such a hypothesis, clearly showing a decreasing SOCs trend as DPSF → DSSF, i.e., at increasing human disturbance.

Multivariate statistics
The FA and PCA combine soil physical-chemical features and SOCs, providing a more complete and complex perspective by quantifying relationship with factors other than human disturbance and depth.Furthermore, these models allow hypotheses about mechanisms involved in SOCs behavior compared to pedoenvironmental features to be explored.

Factor analyses (FA)
Factor 1 (F1) shows that more natural, well-preserved fragments, i.e., the LPSF and LSSF, which feature a higher amount of SOM, thanks to a more developed forest cover and species complexity, are also characterized by (i) higher macronutrient contents (N and P), and (ii) higher CEC, exchangeable cation, and % base saturation, thus reflecting overall higher soil fertility.Under these conditions of high naturality, SOCs increased, with the CHNa method showing a higher statistical correlation than the WBm, confirming its higher analytical reliability.F1 also showed that increasing SOM led to a significant increase in Cu and Mn soil availability, which is consistent with SOM being one of the most influential factors modulating macro-and micro-nutrient availability in tropical soils 2 .Factor F1 can be explained as the influence of BAF naturality in stocking higher amounts of SOC thanks to favorable pedoenvironmental conditions.Factor F2, confirmed the strong relationship between SOM and SOCs, showing that at increasing BAF fragments naturality, we observed an increase in SOM and consequently stocked SOC.This factor adds to the pivotal role played by iron, showing that it increased at increasing SOM amounts.This was due to the well-known processes of SOM accumulation enhanced by organo-mineral interactions in tropical soils, with iron oxides playing an important role in stabilization 73 .Thus, F2 can be explained as the Fe vs SOM interaction.Factor F3 showed that at increasing bulk density (BD), i.e., soil depth (vide supra), a decrease in SOC and consequently soil macro (N), micronutrients (S, B), and exchangeable cations (K+), was observed.Since we know that such an increase in BD along soil depth follows the TFC > DPSF, DSSF > LPSF, LSSF trend, i.e., is inversely correlated to the BAF fragment conservation state, this factor can be explained as the key role of BD in indicating BAF conservation state.This interpretation is consistent with previous studies 74,75 , which showed that mean bulk density increased with soil depth, with this process being particularly enhanced by passing from well-preserved to degraded areas.Pontes et al. 76 argued that human activities increased soil BD while decreasing porosity and creating a hostile edaphic environment for plant roots.Consequently, after human disturbance ends, the regeneration processes will be affected by BD initial conditions and the time required for establishing good edaphic conditions for plant roots.As the first pioneer species started colonization, porosity improved, creating more favorable conditions for hosting species from later successional stages.During this process, which requires a long time depending on starting conditions, BD will decrease, thus indicating improved conditions in the time-dependent steps toward BAF recovery.

Principal component analyses (PCA)
The PCA helps to visualize the FA outcomes.The five investigated BAF fragments are distinctively grouped with the greatest difference between group centroids along the first principal component (Fig. 2).Soil samples collected from the most developed, natural BAF physiognomies, i.e., LPSF and LSSF, form a distinct group along the first principal component.Arrows (indicating soil physical-chemical parameters) showed that these environments correlate with increased SOM and, consequently, SOC stocks and related extraction methods, together with N and B that are typically macro-and micronutrients in well-structured Atlantic Forest pedoenvironments 27 .The worst preserved BAF fragment (TFC) is located on the opposite position of LPSF and LSSF, visually confirming that soil samples collected in this environment have lower levels of all measured parameters besides bulk density.This environment is mainly influenced by Al 3+ and BD, i.e., by acidic and more compacted soils.Again, PCA also confirmed that DPSF/DSSF physiognomies represent a transitional stage between highly-preserved (LPSF and LSSF) and most degraded (TFC) BAF fragments.Samples collected in DPSF and DSSF environments form two distinct groups located along the passage from TFC to LPSF and LSSF.The most important parameters in these environments are mainly soil micronutrients, which are pivotal in enhancing forest development towards more stable conditions 27 .
Overall, soil development and conservation features are strongly influenced by vegetation/forest conditions and vice versa, thus showing a positive feedback relationship (Fig. 2).As human activities (Fig. 2, number "1") with related disturbance end (Fig. 2, letter "a"), after a variable period of time (depending on human disturbance intensity and years passed after last activities), vegetation start a relatively slow process of recovery (Fig. 2, letter "b").Such a process strongly improves (Fig. 2, number "2") soil's overall physical-chemical properties (Fig. 2, letter "c"), including fertility and the amounts of SOC stocked in the entire profile.Indeed, as vegetation reaches higher maturity, canopy cover increases, which results in (a) increased organic material inputs and (b), the formation of a thick soil organic (O) horizon; factors (a) and (b) preserve SOCs from mineralization and leaching processes thus substantially increasing its reserves along the entire soil profile 10 .Indeed, shrub and tree species create more organic complexes and deeper root networks, up to 18/20 m 9,10 .Consequently (Fig. 2, number "3"), the overall soil-plant system improves over time with respect to nutrient exchange, overall fertility, quantity of SOCs, complexity, biodiversity, etc., until reaching a dynamic equilibrium representing the forest/soil climax.The duration to reach this climax is strongly dependent on human disturbance, including the number of years since anthropogenic activity ended and whether additional effects such as climate change affect forest growth and development patterns.

Conclusions
Assessing soil organic carbon stocks (SOCs) is a pivotal step in the challenging research field of monitoring and evaluating GHG emissions associated with the soil-climate feedback cycle.These activities are based on field data collection and laboratory analysis, requiring reliable instruments.A great gap can exist in underdeveloped and developing countries due to resource constraints, and these same countries will experience some of the most severe impacts from climate change, including damages in both environmental and socio-economic terms.Additionally, all of these impacts are enhanced in tropical systems, which play a pivotal role in the global carbon cycle and, consequently, in climate behavior.Brazil represents a paradigmatic example of a previously undeveloped country that, in a relatively short time, completely changed its economy, including dramatic landuse changes, that had both positive and negative, direct and indirect consequences on SOC stocks and behavior.The Brazilian Atlantic Forest (BAF) has been one of the most damaged ecosystems, together with the Amazon, making research on it strategically important worldwide.Soil organic carbon stocks quantification showed that the use of CHN analyzer (CHNa) was highly reliable compared to the Walkley-Black method (WBm).We suggest promoting collaborations between under-equipped laboratories and research groups with high experience with CHNa use to increase network-based research programs, thus providing more reliable data.Additionally, in tropical soils SOCs should be evaluated to at least 1 m depth.A large amount of SOCs is not properly considered if only the first 0-30 cm are collected, which can make further uses of these data (for instance, for GHG emission models) unreliable.Considering investigated BAF physiognomies, representing fragments with different levels and recency in human disturbance, the LPSF and LSSF are the most developed and well-preserved, with soils showing the highest SOC stocks.The Transitional Semideciduous Seasonal Forest to Cerrado (TFC) has the least developed forest canopy and worst preserved soil conditions, thus showing the lowest SOC stocks.The DPSF and DSSF represent a transitional stage among the LPSF/LSSF and TFC, with SOCs values slightly lower than in LPSF/LSSF while higher than in TFC physiognomies.Multivariate statistics showed that the investigated environments feature a complex dynamic feedback where the influence of multiple parameters can play a pivotal

Study area
The research was conducted inside the "Edgardia and Lageado" (EL) experimental farms belonging to the Capivara River basin, Tietê Valley region, Botucatu Municipality in the south-east of Brazil (22° 47′ 30′′-22° 50′ S and 48° 26′ 15′′ to 48° 22′ 30′′ W; Fig. 3).The climate is classified as Cfa (Köeppen criteria 77 ), i.e., subtropical which features hot temperatures (mean annual T: 20.6 °C) and high air humidity (mean annual P: 1707 mm).The sites are located in the geological dominion of the western São Paulo Plateau, which is composed of igneous rocks.The research area is specifically a basaltic cuesta (from Spanish "slope"), which is a geological structure (technically an "homoclinal ridge") with cliff/escarpment on one side and a dip/backslope on the other 78 .
Soils vary from low to medium-high fertility, depending on site overall features 27 .However, most are considered extremely fragile and under threat due to erosion processes enhanced by human activities.From a classification viewpoint 79 , Oxisols (deep Oxic Bo horizon) are the most widespread soil order.They are followed by "young" Entisols, passing to more developed Inceptisols (with a deep diagnostic Cambic Bw horizon).Alfisols and Ultisols (both featured by an Argillic Bt horizon) are present in small areas.All investigated soils had an udic moisture regime 27 .

Experimental design
The vegetation cover greatly varied along the entire "Edgardia and Lageado" experimental farms.Indeed, as with the entire BAF over the last 500 years, the vegetation has been dramatically affected by anthropogenic disturbance and harvest, which is now found in fragmented patches with their own specific physiology and flora associations.For the purpose of classifying BAF fragments into a gradient of human disturbance, we performed an ad-hoc investigation of: (i) historical land-use changes compared using aerial and satellite images; (ii) physiognomic features; (iii) species composition; (iv) phytosociological associations; and; (v) ecological indicators, in addition to review of previous studies 27,[80][81][82] .Based upon this investigation, we grouped sites into five main fragments, ordered from least to most human disturbance: • Late primary semideciduous seasonal forest (LPSF) semideciduous seasonal forest are characterized by the occurrence of a seasonal climate (drought-cold winters-rainy summers) responsible for the foliage's semideciduousness, i.e., partial leaf fall.The term late primary are used for indicating BAF fragments containing almost-native vegetation cover.Indeed, due to their location in not easily accessible areas, they experienced a very low level of human disturbance, with no history of deforestation, selective logging, livestock or agriculture, at least during the last century (Fig. 3a).• Late secondary semideciduous seasonal forest (LSSF) represent a case of advanced regenerative LPSF phase, after a human disturbance (such as logging of forest species for economic purposes), usually occurring more than 30-40 years ago.The advanced regenerative processes (if compared to DPSF and DSSF, below) was favored by their not easily accessible location.• Disturbed primary semideciduous seasonal forest (DPSF) these are BAF fragments containing less natural conditions compared to Late Primary forest (vide supra).Indeed, the forest structure is less developed due to more intensive selective logging at the beginning of the last century.Additionally, these fragments are located in more easily accessible areas.• Disturbed secondary semideciduous seasonal forest (DSSF) represent a case of moderate regenerative DPSF phase.In particular, after a human disturbance such as logging of forest species, fire, and agricultural activities, occurring in a more recent time compared to DPSF.• Transitional semideciduous seasonal forest to Cerrado (TFC) are those BAF fragments affected by a clear decrease in forest canopy cover due to selective logging, grazing, agricultural activities, etc., usually occurring more than 30-40 years ago.Even if the term Cerrado, indicates a specific Tropical savanna ecoregion, in the investigated area, it represents an ecotonal transition between BAF and itself 38 .
Thus, investigated areas are or have been affected by human influence, even if they are now considered "natural" or "seminatural" environments.Indeed, there has been continuous interaction and perturbation by humans throughout the history of the BAF 27 .In fact, there are no BAF fragments that have not been disturbed or influenced by human activities to some degree.Thus, when investigating such fragments, the history of human interaction and activities must be carefully incorporated since humans represent a driver force in the forest's past, present, and future development and behavior 27 .
Eighteen 20 × 100 m (2000 m 2 ) permanent plots (Fig. 3b, Supplementary Material 2) were randomly distributed in LPSF, LSSF, DPSF, DSSF, and TFC.The plot size and distribution were defined according to the methodology proposed by Roder et al. 27 .Specifically, the sampling unit and their distribution along the five investigated BAF fragments were selected after careful in-field study and GIS investigation carried out before the beginning of the research by considering the difference between: (i) the vegetation from several perspectives (vide supra); (ii) morphological and slope aspects; (iii) soil features; (iv) history in human disturbance; (v) area extension of each investigated kind of vegetation.Consequently, the 18 plots were distributed in numbers of 2, 1, 10, 3, and 2 in LPSF, LSSF, DPSF, DSSF, and TFC, respectively (Fig. 3b).Due to the previously reported factor, the uneven distribution was necessary to capture the whole variance among investigated environments.Milliren et al. 83 assessed whether uneven plot distribution can bias random effect estimation and, consequently, observed ) clearly confirmed such a behavior, showing a strong positive correlation between pH vs SOM (r = 0.62**), CEC (r = 0.53**) and, consequently, all basic cations (K + , r = 0.61**; Ca 2+ , r = 0.70***; Mg 2+ , r = 0.72***); and an inverse correlation was observed for SOM vs Al 3+ (r = − 0.35**).

Figure 1 .
Figure 1.Biplot of a principal component analysis.

Table 2 .
Soil S and micronutrients content in the investigated BAF's fragments (mean ± SE).Legend as explanations as in Table1.

Table 4 .
SOC content according to investigated BAF fragment, soil depth, and applied method.Legend as in Table1.Different capital letters show a significant difference (p < 0.05) among investigated BAF fragment; different lowercase letters show a significant difference (p < 0.05) among investigated soil depth.
role in determining differences and commonalities among recovering forests.Regarding management practices, we demonstrated that BAF requires a long time to fully recover its initial conditions after the end of human disturbance; consequently, preservation and conservation strategies should be privileged in remnant fragments.