Maracá is a riverine island located in the Rio Uraricoera in Roraima State (Brazil) and has an area of about 100,000ha. A forest type dominated by Peltogyne gracilipes Ducke (Caesalpiniaceae) occurs on Maracá Island on a range of soil types. This study compares the structure, floristic composition and soils of the Peltogyne forest with the most widespread lowland forest type on Maracá. Three 0.25ha plots were set up in each of three forest types: Peltogyne-rich forest (PRF), Peltogyne-pooi forest (PPF) and forest without Peltogyne (FWP). Within each plot all trees (≤ 10cm dbh) were recorded. Seedlings and saplings were sampled in subplots of 2m × 1m (seedlings) and 4m × 4m (saplings). In the PPF and FWP, Sapotaceae were the most important family with the highest relative dominance and relative density values. Caesalpiniaceae showed high values in the PRF and PPF. Licania kunthiana, Pradosia surinamensis and Simarouba amara occurred in the canopy layer in all the forest types. Peltogyne dominated the canopy in the PRF and comprised 20% of stems and 53% of the total basal area of all trees ≥ 10cm dbh, and 91% of the stems and 97% of the total basal area of individuals ≥ 50cm dbh. In PPF, Lecythis corrugata and Tetragastris panamensis were the most abundant species, followed by Peltogyne. In the FWP the most abundant trees (≥ 10cm dbh) were Licania kunthiana and Pradosia surinamensis. In all forest types, the soils were sandy and acid with low concentrations of extractable phosphorus and exchangeable cations, but the soils under PRF were notably richer in magnesium.

The Brazilian Atlantic Forest (AF) covers 13% of Brazil but retains only 26% of its original forest area. Utilizing a Morphological Spatial Pattern Analysis (MSPA), we generated 30 m spatial resolution fragmentation maps for old-growth and secondary forests across the AF. We quantified landscape fragmentation patterns and carbon (C) dynamics over 35 years using MapBiomas data between the years 1985 and 2020. We found that from 1985 to 2020 the forest suffered continuous fragmentation, losing core (nuclei forest fragments) and bridge (areas that connect different core areas) components of the landscape. About 87.5% (290 468.4 km²) of the remaining forest lacked core areas, with bridges (38.0%) and islets (small, isolated fragments) (35.4%) being predominant. Secondary forests (1986–2020) accounted for 99 450.5 km² and played a significant role in fragmentation pattern, constituting 44.9% of the areas affected by edge effects (perforation, edge, bridge, and loop), 53.7% of islets, and comprising only 1.4% of core forest. Additionally, regeneration by secondary forests contributed to all fragmentation classes in 2020. Even with the regrowth of forests, the total forested area in the biome did not increase between 1985 and 2020. Deforestation emissions reached 818 Tg CO₂, closely paralleled by edge effects emissions at 810 Tg CO₂, highlighting a remarkable parity in C emissions between the two processes. Despite slow changes, AF biome continues to lose its C stocks. We estimated that around 1.96 million hectares (19 600 km²) of regenerated forest would be required to offset the historical C emissions over the analysed period. Hence, MSPA can support landscape monitoring, optimizing natural or active forest regeneration to reduce fragmentation and enhance C stocks. Our study’s findings are critical for guiding land-use policies focusing on minimizing emissions, promoting forest regrowth, and monitoring its permanence. This study offers biome scale, spatially explicit information, critical for AF conservation and management.