Dynamics of Extreme Floods in Southeast and South Brazil

Abstract:

Many extreme floods result from a causal chain, where exceptional rain and floods in water basins from different sizes are related to large scale, anomalous and persistent patterns in atmospheric and oceanic circulation. Organized moisture plumes from oceanic sources are often implicated. One could use an Eulerian-Lagrangian climate model to test a causal chain hypothesis, but the parameterization and testing of such a model covering convection and transport continues to be a challenge. Consequently, empirical data based studies can be useful to establish the need to formally model such events using this approach. Here we consider two flood-prone regions in Southeast and South Brazil as case studies. A hypothesis of the causal chain of extreme floods in these regions is investigated by means of observed streamflow and reanalysis data and some machine learning tools. The signatures of the organization of the large scale atmospheric circulation in the days prior to the flood events are evaluated based on the integrated moisture flux and its divergence field and storm track data, so that a better understanding of the relations between the flood magnitude and duration, strength of moisture convergence and role of regional moisture recycling or teleconnected moisture is established. Persistent patterns and anomalies in the sea surface temperature (SST) field in the Pacific and Atlantic oceans that may be associated with disturbances in the atmospheric circulation and with the flood dynamics are investigated through composite analysis. Finally, machine learning algorithms for nonlinear dimension reduction are employed to visualize and understand some of the spatio-temporal patterns of the dominated climate variables in a reduced dimensional space. Prospects for prediction are discussed.