The slow downwearing of Madagascar: Inferring patterns and controls on long-term basin-averaged erosion rates from in situ Be-10 at the catchment and regional level

Cosmogenic nuclide analysis of river sediment provides insight into catchment-wide erosion rates and dynamics. Here, we investigate spatial patterns and controls on Be-10-inferred erosion rates in Madagascar, a moderately seismically active microcontinent surrounded by passive margins with locally steep topography and a climate that varies from humid tropical to semiarid. We use a compiled dataset of 99 detrital Be-10 measurements, 63 of which are new, covering more than 30% of the country and a wide range of topographic, bioclimatic and geologic characteristics. Overall, Be-10 erosion rates are low (2.4-51.1 mm kyr(-1)), with clear differences between regions. The lowest rates are measured on the central highlands (similar to$$ \sim $$ 8 mm kyr(-1)), in the Alaotra-Ankay graben (similar to$$ \sim $$ 11 mm kyr(-1)) and in the large north-central catchments (similar to$$ \sim $$ 11 mm kyr(-1)). Higher rates are found on the steep eastern escarpment (similar to$$ \sim $$ 20 mm kyr(-1)), in the northwest (similar to$$ \sim $$ 31 mm kyr(-1)) and in the southwest (similar to$$ \sim $$ 29 mm kyr(-1)). A stepwise linear regression model identified elevation as the main factor associated with variations in Be-10 erosion rates (lower rates for higher catchments). Random within-between statistical models (REWB), on the other hand, indicated that the differences between different regions can be explained by differences in river concavity, seismic events and gully (lavaka) densities, whereas additional variation within regions is only linked to seismicity. We find no correlation between catchment or river steepness and Be-10-inferred erosion rates. Our results indicate that in Madagascar, long-term erosion rates are overall low and that simple topography-based models do not explain variations in rates of landscape change inferred from Be-10 concentrations in river sediment. We demonstrate that identifying different regions aids in interpreting spatial patterns of erosion rates and that REWB models can be a powerful tool in deciphering environmental controls on Be-10 erosion rates.