Impact of climate and volcanism age on landscape evolution of oceanic islands

Tropical hotspot volcanic islands are exceptional laboratories for understanding climate-landscape interactions. Réunion Island, in the southern Indian Ocean, is particularly interesting because it has one of the largest rainfall gradients in the world: from 0.5 to 11 m/yr over a 50 km distance. It experiences frequent cyclones, such as the January 15^th, 2024 cyclone Belal that produced from ~0.1 to ~1 m of rainfall in 24 hours. Réunion is composed of two shield volcanoes: Piton des Neiges (3070 m, dormant) and Piton de la Fournaise (2632 m, active). Here, we focus on long-term (Myr) and short-term (<yr) erosion of Réunion Island.

Long-term basin-averaged erosion rates based on excavated lava volumes of 29 catchments located on both the leeward and windward side of the island range from ~10^-3 to ~10 mm/yr. Erosion rates for basins eroding < ~1 mm/yr show a positive relationship with mean annual precipitation. They also have a negative relationship with the duration of erosion, i.e. the age of the incised volcanic surface, which we consider as a proxy for post-eruption basin aging. Fluvial incision parameters recovered from numerical modelling of the stream power law show a positive trend with mean annual cyclonic rainfall.

To disentangle the dual effect of climate and post-eruption surface aging on erosion, we perform a series of correlation analyses between mean rainfall and rainfall variability, river discharge, catchment morphometrics, and time elapsed since the latest eruptions for 136 basins. We find that Hack’s law exponent (the relationship between basin area and river length), drainage density, and large basins width and hypsometry integral follow a temporal trend (mostly on Piton des Neiges), which confirms published work on stratovolcanoes. Discharge variability has no apparent relationships with basin geometry and erosion. Interestingly, Piton des Neiges and Piton de la Fournaise volcanoes have different main climate erosion drivers: the first one seems to be more influenced by mean annual precipitation, and the second by precipitation variability.

To test whether the effect of rainfall variability and mean annual rainfall on erosion, which we observe at long timescales, can be detected at short timescales, we conducted three field surveys of river channels in October 2023, February 2024, and October 2024, i.e. before and after cyclone Belal. We used photogrammetry to survey the river sediments and to extract 3D shapes and grain size distributions of channel bedload (pebble, cobble, boulder) from point clouds. In addition, we assessed local channel modifications, including sedimentary deposition and excavation, terrace undercutting, and movement of large boulders.  These data will provide erosion thresholds for fluvial incision, and will allow comparative temporal and spatial analysis of grain sizes distribution. In the future, more investigation will be needed on denudation at thousand-year timescales to bridge the gap between our short- and long-term studies.