Abstract. The extent to which the central highlands of Madagascar
were once covered by forests is still a matter of debate: while
reconstructing past environments is inherently difficult, the debate is
further hampered by the fact that the evidence documenting land cover
changes and their effects on carbon and sediment dynamics in Madagascar has
hitherto mainly been derived from lake coring studies. Such studies provide
an integrated view over relatively large areas but do not provide
information on how land-use change affects hillslopes in terms of carbon and
sediment dynamics. Such information would not only be complementary to lake
inventories but may also help to correctly interpret lake sediment data.
Carbon stable isotope ratios (δ13C) are particularly useful
tracers to study the past dynamics of soil carbon over time spans ranging
from years to millennia and thus to understand the consequences of land-use
change over such time spans. We analysed soil profiles down to a depth of 2 m
from pristine forests and grasslands in the Lake Alaotra region in central
Madagascar. Along grassland hillslopes, soil organic carbon (SOC) content
was low, from 0.4 % to 1.7 % in the top layer, and decreased rapidly to ca.
0.2 % below 100 cm depth. The current vegetation predominantly consists
of C4 grasses (δ13C ∼ −13 ‰), yet topsoil δ13C-OC ranges between −23.0 ‰ and −15.8 ‰, and most profiles show a decrease in δ13C-OC with depth. This contrasts with our observations in the
C3-dominated forest profiles, which show a typical profile whereby δ13C values increase slightly with depth. Moreover, the SOC stock of
grasslands was ∼ 55.6 % lower than along the forested
hillslopes for the upper 0–30 cm layer. δ13C values in
grassland and forest profiles converge to similar values (within 2.0 ± 1.8 ‰) at depths below ∼ 80 cm,
suggesting that the grasslands in the Lake Alaotra region have indeed
developed on soils formerly covered by a tree vegetation dominated by C3
plants. We also observed that the percent of modern carbon (pMC) of the bulk
OC in the top, middle and lower middle positions of grasslands was less than
85 % near the surface. This could reflect a combination of (i) the long
residence time of forest OC in the soil, (ii) the slow replacement rate of
grassland-derived OC (iii) and the substantial erosion of the top positions
towards the valley position of grasslands. At the valley positions under
grassland, the upper 80 cm contains higher amounts of recent grass-derived
OC in comparison to the hillslope positions. This is likely to be related to
the higher productivity of the grassland valleys (due to higher moisture and
nutrient availability), and the deposition of OC that was eroded further upslope
may also have contributed. The method we applied, which is based on the
large difference in δ13C values between the two major
photosynthetic pathways (C3 and C4) in (sub-)tropical terrestrial
environments, provides a relatively straightforward approach to
quantitatively determine changing vegetation cover, and we advocate for its
broader application across Madagascar to better understand the island's
vegetation history.