No abstract

The possibility of ecosystem boundary changes in northern Brazilian Amazonia during the Holocene period was investigated using soil organic carbon isotope ratios. Determination of past and present fluctuations of the forest-savanna boundary involved the measurement of natural C isotope abundance, expressed as δC, in soil organic matter (SOM). SOM C analyses and radiocarbon dating of charcoal fragments were carried out on samples derived from soil profiles taken along transects perpendicular to the ecotonal boundary. SOM δC values in the upper soil horizons appeared to be in equilibrium with the overlying vegetation types and did not point to a movement of the boundary during the last decades. However, δC values obtained from deeper savanna and forest soil layers indicated that the vegetation type has changed in the past. In current savanna soil profiles, we observed the presence of mid-Holocene charcoals derived from forest species: fire frequency at that time was probably greater, and more extensive savanna may have resulted. Isotope data and the presence of these charcoals thus suggest that the forest-savanna boundary has shifted significantly in the recent Holocene period, forest being more extensive during the early Holocene than today. During the middle Holocene, the forest could have strongly regressed, and fires appeared, with a maximum development of the savanna vegetation. At the beginning of the late Holocene, the forest may have invaded a part of this savanna, and fires occurred again.

No abstract

No abstract

The impact of deforestation on soil organic carbon (SOC) stocks is important in the context of climate change and agricultural soil use. Trends of SOC stock changes after agroecosystem establishment vary according to the spatial scale considered, and factors explaining these trends may differ sometimes according to meta-analyses. We have reviewed the knowledge about changes in SOC stocks in Amazonia after the establishment of pasture or cropland, sought relationships between observed changes and soil, climatic variables and management practices, and synthesized the δ C measured in pastures. Our dataset consisted of 21 studies mostly synchronic, across 52 sites (Brazil, Colombia, French Guiana, Suriname), totalling 70 forest-agroecosystem comparisons. We found that pastures (n = 52, mean age = 17.6 years) had slightly higher SOC stocks than forest (+6.8 ± 3.1 %), whereas croplands (n = 18, mean age = 8.7 years) had lower SOC stocks than forest (-8.5 ± 2.9 %). Annual precipitation and SOC stocks under forest had no effect on the SOC changes in the agroecosystems. For croplands, we found a lower SOC loss than other meta-analyses, but the short time period after deforestation here could have reduced this loss. There was no clear effect of tillage on the SOC response. Management of pastures, whether they were degraded/nominal/improved, had no significant effect on SOC response. δ C measurements on 16 pasture chronosequences showed that decay of forest-derived SOC was variable, whereas pasture-derived SOC was less so and was characterized by an accumulation plateau of 20 Mg SOC ha after 20 years. The large uncertainties in SOC response observed could be derived from the chronosequence approach, sensitive to natural soil variability and to human management practices. This study emphasizes the need for diachronic and long-term studies, associated with better knowledge of agroecosystem management.