Potential reconstruction of 19th century flood variability in the northeastern Amazon using tree-ring δO18

The future of the Amazon basin is of great concern front projected impacts due to climate change. Current estimates of hydrological changes are said to point toward unprecedented stages in many aspects, including floods and droughts. However, hydrological monitoring is often limited in time and can mask long term natural variability, affecting conclusions regarding present and future trends. Therefore, tree-ring time series are a valuable complement to this kind of assessment and can provide insights on the long-term occurrence of small and large floods and droughts. Here, we propose a preliminary reconstruction of annual floods from 1850 to 2016 based on a tree-ring δO18 series for the Paru basin, located in the northeastern Amazon. Isotope data present strong (r between 0.6-0.9 in module) and spatially consistent correlation with annual floods and with basin-aggregated rainfall from 1980-2016 at the Paru and nearby basins. This encouraged us to explore a simple linear regression model by fitting δO18 and annual flood series. The model was fitted using simulated discharge from MGB-SA hydrological model from 1980 to 2016 (and compared with observed record), and resulted in a r² > 0.5 for more than 200 river reaches near the tree-ring data’s site. Regression models presented a success rate of >70% in classifying small flood years, while presenting >60% and >40% for regular and large floods respectively. This preliminary assessment indicates the potential of hydrological reconstruction of floods based on Paru’s δO18 data, enabling valuable insights on part of the Amazon hydrological variability since mid-19^th century. Future perspectives could include hydrological modelling based on rainfall ensembles built from this series for more detailed assessments.