Study sites
We studied 12 second-growth forests established for 11-47 years on former pasturelands, 13 restoration plantations of 10-61 years of age established on pasturelands and croplands, and five old-growth, reference forests (Fig. S1; Table S1). Second-growth forests were sampled in the Corumbataí watershed, southeastern Brazil (Fig. S1). Pastureland is the principal land use in this watershed (43.7% of area), followed by sugarcane fields (29.4%), and remaining native forest cover (12.4%; Silvio F. B. Ferraz et al., 2014). Previous land use and the age of regenerating forests were determined based on panchromatic aerial photographs and annual LANDSAT 5 and 8 images. Further information about this watershed and forest patch sampling can be found in César et al. (2017). Second-growth forests of the Corumbataí watershed are found regenerating almost exclusively in marginal lands, such as steep slopes and/or sandy or rocky soil, where mechanized agriculture is not possible. Extensive cattle ranching or eucalypt woodlots have been established in these marginal agricultural lands, and their less intensive historical soil management, at least compared to sugarcane or other crop production, have led to the regeneration of secondary forests when land is abandoned (César et al., 2017). All second-growth forests were found away from riparian areas and were expansions of existing forest remnants.
Restoration plantations were mostly identified in the same region (< 100 km) of second-growth forests, but some additional plantations (three out of the four plantations older than 25 years old) were sampled (< 300 km) to include older sites in our dataset (Fig. S1). These plantations were established with a high diversity of native tree species (20-100 species) and had comparable levels of tree richness than the studied second-growth and reference forests (Fig. S2). Plantation management consisted of planting nursery-grown seedlings in regular spacing (usually 3 x 2 m), fertilizing the soil before (base fertilization) and after planting (broadcast fertilization), and weeding ruderal plants with glyphosate spraying or mowing (see details in P.H.S.; Brancalion et al., 2019). Land use prior to plantation establishment was determined by interviewing restoration project managers. Only three sites were established on pasturelands, and the rest were established on former sugarcane fields. Twelve sites were found in riparian buffers and were established to comply with the Forest Code (Pedro H. S. Brancalion, Garcia, et al., 2016). Nine of the 12 sites were isolated from native forest remnants, and three sites were established within the neighborhood of existing degraded remnants. Consequently, our analysis did not allow for a controlled comparison of properties between tree plantations and natural regeneration because these restoration approaches were established in quite different biophysical conditions in the studied landscapes. Although not appropriate for objectively comparing these restoration approachesper se and isolating uncontrolled factors (Reid et al., 2018), our study design depicts the cost-effectiveness of restoration approaches currently being implemented in agricultural landscapes of Brazil’s Atlantic Forest (see Pedro H. S. Brancalion, Schweizer, et al., 2016).
We complemented our sampling with five reference forests, represented by old-growth remnants distributed around the Corumbataí watershed. These remnants were selected for being the best conserved forests of the study region, with a well-developed forest structure and some large (> 50 cm diameter at breast height) late-successional remnant trees (Fig. S1). These forests do not represent the pre-disturbance carbon stocks of the original forests of this region, since these remnants are small (<200 hectares) and are embedded within a patchy agricultural matrix. However, they represent an appropriate benchmark for restoration in this region, given that restored forests will likely persist as small patches surrounded by agriculture.
The forests sampled were selected according to different criteria. The second-growth forests were selected systematically based on available information on second-growth forest age and previous land use, aiming at a representation of approximately five forests per age class x previous land use (S. F. B. Ferraz et al., 2014). Restoration plantations and old-growth forests were selected based on the availability of these forest typologies with different ages within and closer to the Corumbataí watershed, the focus area of the study. We sampled all restoration plantation >15 years old and old-growth forests we knew in the region due to the scarcity of these kinds of forest, and seven restoration plantations <15 years old based on their location (forests found within the watershed or closer to it was prioritized) and access.