Study sites
This study was conducted at two different areas located in separate ecoregions of the São Paulo State, Brazil. The first area was within the Brazilian Atlantic Forest in a Seasonal Semi-deciduous Forest (SSF), also known as tropical seasonal forest, located inside the borders of the Caetetus Ecological Station. The second area was within the Cerrado ecoregion in an area classified as Cerrado sensu-stricto (CSS) located on the Mogi Guaçu Biological Reserve. In each area two strata were chosen for stratified systematic sampling.
The forest at the Caetetus Ecological Station constitutes one of the most significant remaining areas of SSF in Brazil. The “semi-deciduous” denomination refers to the seasonal cold weather and the reduced water availability in the soil, alongside other environmental factors. As such, most of the arborescent species lose their leaves in winter to reduce water consumption and decrease their rates of growth (Tabanez et al. 2005). According to the Management Plan, the Conservation Unit covers the cities of Galia and Alvinlândia with a total area of 2178 ha. Its elevation ranges from 520-680 m above sea level, and it is relatively flat (less than 6%). It is located on the northern border area of the hydrographic basin of the Paranapanema river. The climate according to the Köeppen (1948) classification is “Cwa”: mesothermal dry winter with temperatures below 18 ° C and above 22 ° C in summer; total precipitation in the driest month is only 30 mm, and total annual precipitation is between 1100 and 1700 mm (Tabanez et al. 2005). In this area, stratum I (SSF1) and stratum II (SSF2) were selected.
The SSF1 stratum was characterized by arboreal vegetation of large stature with sparse herbaceous and graminoid vegetation. SSF1 contained predominantly deciduous species and has experienced anthropogenic disturbance; it occupies the interfluves and the tops of plateaus with a total area of 777.8 ha (area “L" in the Management Plan map). The SSF2 stratum was an area containing dense arboreal vegetation of large stature with a smaller number of deciduous species. Overall, SSF2 was a well-preserved forest occupying high escarpments and plateau edges across a total area of 439.4 ha (area “M" in the Management Plan map) (Tabanez et al. 2005).
The Mogi-Guaçu Biological Reserve, the location of the cerrado field site, is part of the old Campininha Farm located in the Martin Prado Junior district of the city of Mogi Guaçu in the São Paulo State. With a total area of 470 hectares, the landscape is relatively flat (less than a 15% slope). In the area of the Campininha Farm, the elevation ranges from 566 to 724 meters above sea level. The climate is mesothermal with two well defined seasons: a dry winter (from the months of April to September) and a hot summer (from October to March). The annual average rainfall is 1335 mm, and the average temperature is 20.5 ° C (São Paulo 2015). The unit is divided into two main sections: one designated as area “A” and the other as area “B”.
Area “A”, where this research was conducted, covers 343.4 ha and is dominated by cerrado ecosystems at different regeneration stages, as the area has suffered perturbations like frequent fires and occupation with cattle farming (Mantovani and Martins 1993). With the largest area being a Cerrado sensu-stricto (savanna woodland) formation (spanning from pioneer to advanced successional stages); there are also small patches of herbaceous vegetation near the watercourse. Two strata were also selected in this area: CSS1 and CSS2. The CSS1 stratum is characterized as a forested savannah in an advanced successional stage with an area of 136.5 ha; and the CSS2 stratum is a forested savannah in an intermediate successional stage with an area of 197.5 ha (São Paulo 2015).
Field measurements and sample collection
In the field survey, dead woody debris was classified into size classes according to the Keller et al. (2004) classification: small (branches or bamboo with diameter between 2 and 5 cm); medium (branches or bamboo with diameter between 5 and 10 cm); and large (woody material with diameter greater than 10 cm). In this study, only the large elements were measured. Data were collected according to the line intersect sampling (LIS) method, in which an element of downed dead wood was tallied if it was completely intersected by a transect or if it intersected the front end of a transect, following the protocol detailed by Gregoire and Valentine (2008). A professional compass (Suunto KB-20) was used to alignment the transects along the cardinal directions. A 50 m tape measure was used to lay out the transects, with corrections for slope when necessary. Additional details about the field protocols can be found in Moreira (2017) and Moreira et al. (2019).
In each stratum of each area, ten sampling units (Fig. 1) were installed according to a systematic sampling protocol, with 300 m distance between the center points of each sampling unit (point C - Fig. 1). Each sampling unit spans 650 m of transect segments: one 200 m transect segment in the North-South direction (segment AE - Fig. 1); one 200 m transect segment in the East-West direction, crossing the center point of the N-S transect (segment FG - Fig. 1); a 150 m segment crossing the N-S segment in the East-West direction at 25 m south of the center point (segment HI - Fig. 1); and a 100 m East-West segment crossing the N-S segment at 50 m south of the center point (segment JK - Fig. 1).
A caliper was used to measure each element’s diameter, perpendicular to the central axis, at the point where the segment intersected the element. If bark remained attached to the element, the diameter was measured with it. The length of the element was recorded using a tape measure along the central axis of the element (regardless of its shape). Furthermore, each element’s width perpendicular to the transect segment direction was measured according to Gregoire and Valentine (2008). Each element was classified into one of five decay classes according to Harmon et al. (1986). Decay class 1: woody material consisting of solid wood with leaves and/or fine twigs attached to the principal part without noticeable degradation; decay class 2: solid wood material with intact bark but no leaves or fine twigs; decay class 3: solid wood material similar to class 2 except with rotting bark; decay class 4: partially rotten material that can be broken when kicked; and decay class 5: material that is rotten, friable, and can be broken with bare hands.
Though species identification for debris material is difficult, this data was collected for CWD elements whenever possible – most frequently when the presence of bark and attached branches exhibited particular characteristics of certain species (e.g., a specific odor). For each element, a disc sample was collected at the transect intersection point, perpendicular to the central axis. These samples were taken using either a chain saw or a manual saw (for elements that had a high degree of decomposition). Samples were packaged, tagged, and carefully transported to the laboratory.
Laboratory analysis
For density calculations from disc samples, the cylinder extraction method proposed by Keller et al. (2004) was used, with the only variation in this study being that the cylinders were removed in a laboratory setting. The extraction of the cylinders was performed using a bench-mounted drill with a hole saw that produced samples of known diameter. For elements with a high level of decay, sample cylinders were removed manually using an aluminum cylinder. The length of the cylinders was measured and confirmed with a digital caliper; although the hole saw (drill bit) had a known volume, the length of the sample cylinders it produced varied according to the width of the disc brought from the field. The sample cylinders were oven-dried at 65 ° C to a constant weight. Dry cylinders were removed from the oven and weighed to four decimal places on an analytical scale to compute final dry weight.
The sample cylinders from each disc were milled in a knife mill (Wiley type), and sieved through a no. 40 mesh, with equipment being vacuum-cleaned between samples to avoid contamination. Carbon concentration measurements were achieved with LECO-C632 carbon combustion and analysis equipment; i.e., samples are placed in a ceramic chamber that reaches a temperature of approximately 1000 ° C for about one minute, undergoing combustion in the presence of oxygen. The LECO system uses infrared sensors to detect elemental carbon in the outflow of gas from the combustion system, and results are delivered as percentage mass values. For each sample, two replicates were made to determine final carbon concentrations. All laboratory analyses were performed in the Department of Forest Sciences at the University of São Paulo (ESALQ/USP).
Data analysis
The data from the sampling units (650 m of transect segments) were used to calculate wood density and carbon concentration. The wood density (ρk) of each kth element was computed as the average of the sample cylinders’ densities, calculated as the dry weight (g) divided by fresh volume (cm3); average densities were computed for each decay class. For the volume calculation, discs with hollow sections were photographed, and the hollow areas were measured using ImageJ (Rasband 2016). Thus, the volume for each element (vk) was calculated as: the cross-sectional area of the kth element minus the hollowed cross-sectional area of the kth element then multiplied by the length of the kth element.
The biomass (bk) value for each kth element was determined as the product of the element’s volume and its density: bk=Vk.ρk. The carbon stock in each kth element was calculated as the product of its biomass and the carbon concentration average computed via the laboratory analysis by LECO-C632 equipment then averaged according to decay class. From the SSF area, n=506 elements were analyzed for density d carbon concentration; from the CSS area, n=182 elements were analyzed.
From the sampling unit scheme, Moreira et al. (2019) tested different configurations and segments length, selecting the design based on the estimates with the smallest standard error, narrowest confidence interval, and lowest relative error. The cross-shaped with 150 m segment length (segments BE and HI - Fig. 1) was selected for CSS type, and cross-shaped with 200 m segment length (segments AE and FG - Fig. 1) for SSF.
The total estimation of biomass/carbon stock per hectare by sampling unit s\(\left (\hat {\lambda }_{y\pi s}\right)\) was calculated using the estimator \(\hat {\lambda }_{y\pi s}^{c}=\sum _{U_{k} \in s} \frac {y_{k}} {w_{k}(\theta _{s})\,L}\), where L is the length of both segments that makes the cross shaped design, yk is the parameter of interest (biomass or carbon stock), and wk is the perpendicular width to the transect direction, (θs). The estimation for each unit was replicated, and all estimates were combined for the whole population. See details about the estimators in Moreira (2017), Moreira et al. (2019), as well as Gregoire and Valentine (2008). All data were processed and analyzed using R (R Core Team 2018).