Study site
Forests cover approximately 35% (356,426 ha) of the overall administrative territory of the Basilicata region (Fig. 1). Such forest surfaces are spatially fragmented and distributed within a broad altitude range, from approximately 300 to 1700 m a.s.l. Less than 50% of these surfaces are public property and are owned and managed by land administration agencies (i.e., municipalities, national, and provincial park institutions). At the same time, the remaining forests are private properties owned by citizens or families rather than private companies. The size of forest surfaces varies according to property ownership, with public forests consistently over 50 ha. In these forests, the dominant vegetation categories are represented by broadleaved forests, with an evident prevalence of Mediterranean deciduous oaks covering 51.8% of the total forest area, followed by European beech (8.4%), Mediterranean evergreen oaks (7.9%), thermophilous shrublands (6.9%), other mesophilic and mesothermophilic broad-leaved forest stands (5.5%), riparian forests (3.9 %), chestnut forests (3.0%), and plantation forests with exotic tree species (0.6%) (Costantini et al. 2006). Conifer forests cover the remaining 12% of the forest area, of which subalpine and alpine conifers represent 3% and Mediterranean pine forests cover 9%.
Collection of forest management plans and stand attributes
For each municipality, we acquired the forest management plans (FMP) in force for the public forest surface under a management plan. The FMP represents a pivotal document that provides forest resource inventory and spatiotemporal and quantitative management information at the forest parcel scale. The parcel serves as the smallest management unit of forest stands in which both ecological and silvicultural management criteria are implemented. From each FMP, we extracted an array of topographic and biometric attributes to quantitatively characterize public forest surfaces. At the parcel scale, the following inventory attributes were extracted: (i) surface area of the forest parcels (ha); (ii) the mean annual increment (m3 ha−1); and (iii) the aboveground standing volume (m3 ha−1). Moreover, the forest parcel polygons were used as spatial vector data. For each centroid of the forest parcel, the altitude value (m a.s.l.) was extracted from the freely available digital elevation model (DEM) at a resolution of 20 m (http://www.sinanet.isprambiente.it/it/sia-ispra/download-mais/dem20/view).
Estimating the volume of forest residues
Commonly, the term “forest biomass residues” refers to woody biomass generated directly by forest management activities, such as logging residues, and by the wood processing industry, such as shavings, sawdust, and woodchips. Although biomass residues are generated from forest logging activities to the roundwood processing plant, in this study, we considered exclusively the biomass residues produced by on-site harvesting activities. Therefore, in our study, the term forest residues refers to the biomass residues directly resulting from the harvesting and processing activities of on-site logged trees. Such residues are defined as the primary class of forestry-derived biomass feedstock obtainable from forest harvesting, which is also indicated with the common terms “slash” or “brash” (Titus et al. 2021) and more specifically as fine wood debris (FWD; Camia et al. 2021).
Forest biomass residues mainly consist of upper stem and branch fractions with a diameter lower than 5 cm, excluding the stump, roots, and leaf biomass (except for conifers) of harvested trees. We quantified forest residues as a fraction of the harvestable aboveground standing volumes of each parcel of FMPs. Then, according to the species-specific percentage values estimated by Cozzi et al. (2013), for Basilicata forests, the volume of forest residues was estimated as 9 to 20% of total forest utilization (i.e., harvested aboveground standing volume, m3) of each parcel.
Estimating the carbon content of biomass residues
To assess the organic C fraction potentially left on site for each forest parcel (x), the quantity of organic C content by forest residues (ROC, Mg ha−1) was calculated as follow:
$$ {\mathrm{ROC}}_x={\mathrm{AV}}_x\bullet \mathrm{c} $$
where, AVx represents the aboveground residue volume (m3 ha−1) at parcel scale, and c is the conversion factor from volume to weight of organic C (Mg m−3). Since c depends on the tree species, for oak woods, beech woods, chestnut woods, hornbeam forests, other deciduous forests, and conifers, the species-specific c values were retrieved from Gasparini et al. 2013.
Estimation of soil organic carbon stock (SOCS)
The SOC stock (Mg C ha−1) was calculated from the SOC content, bulk density, and coarse fragments at six depth intervals (0–5 cm, 5–15 cm, 15–30 cm, 30–60 cm, 60–100 cm, and 100–200 cm). Finally, the value of SOC at each parcel scale was obtained by summation of the organic contents of all six soil depth intervals. The stock of SOC data (OCSTHA in the SoilGrids database) was obtained from the recently released International Soil Reference and Information Center (ISRIC) (Hengl et al. 2017). The data available in a gridded format at a spatial resolution of 250 m were downloaded from the server ftp://ftp.soilgrids.org/data/recent.
Scenarios of in situ released biomass residues
To assess the potential availability of biomass residues for the Basilicata region, different biomass-releasing scenarios were simulated by computing them on the total available harvesting residues at the parcel scale (Fig. 2).
It was simulated from the release of 100% of the available forest residues as a baseline scenario. This scenario synthesizes a set of socio-economic, technological, and environmental circumstances, which justify the retention of the forest biomass residues at a logging site. Therefore, such a baseline scenario includes leaving all forest residues on site or processing them so that they are mulched or distributed on site for forest sustainable management goals, excluding any commercial and industrial use.
Three different scenarios were simulated by gradually decreasing the release of forest biomass residues from 100% to 70%, 50%, and 30%. These decreasing release scenarios consist of leaving a progressively lower quantity of biomass residues on forest sites, excluding them from commercial and industrial uses. These scenarios reflect increasing interest in the large-scale utilization of forest biomass residues for bio-based industry processes by quantifying their potential availability at a regional scale. Each scenario synthesizes a set of favorable socio-economic, technological, and environmental circumstances, allowing the removal of biomass residues from logging sites for commercial and industrial purposes. Moreover, the 30% scenario release represents the minimum benchmarking fraction of residues that should be left on site according to the guidelines for forest residue release in northern European countries (Titus et al. 2021). Therefore, we used 30% as a conservative threshold to minimize soil fertility loss in Mediterranean mountain ecosystem forests.
Biomass residue sustainability ratio
For each simulated scenario, the sustainability of forest residue removal was evaluated by means of the ratio between the ROC and the existing SOC contents as follows:
$$ {\mathrm{RC}}_{\mathrm{Index}}=\frac{\mathrm{ROC}}{\mathrm{SOC}} $$
where RCIndex corresponds to the residue C index, and ROC and SOC correspond to the C content of the forest residues and soil, respectively. The range values of the dimensionless RCIndex are > 0; when its value is close to zero, the releases of forest residues on logging forest floor are low or null, while it assumes increasing values > 0 when the amount of forest residues on site increases.
Data analysis
To identify the group of forest parcels with similar growth and productive attributes, an unsupervised K-means clustering analysis was performed (Hartigan and Wong 1979). The clusters of the forest parcel were partitioned by site and stand attributes: altitude, SOC, mean annual increment (MAI), and aboveground standing volume (ASV). The K-means algorithm allows identification of the cluster of forest parcels characterized by high intra-class similarity. The Euclidean distance method was used to assign each forest parcel to the closest cluster centroid. The optimal number of clusters was determined by the gap statistic (k) iterative method with 500 Monte Carlo bootstrapped samples, and by employing the first SE max criteria (with SE factor of 1) to identify the location of the maximum gap statistic value.
The contribution of each attribute to both dimensions was determined by principal component analysis (PCA). In the K-means partitioning clustering method, PCA is used to reduce the dimensions of the data. Each variable was considered significant when its percentage contribution exceeded the expected average contribution.
Management and analysis of the data were performed in R (R Core Team 2020) by means of “cluster” (Maechler et al. 2021), “FactoMineR” (Lê et al. 2008), and “factoextra” (Kassambara and Mundt 2020) packages.