Thesis Defense
Student: Camila Ribeiro Sapucci
Program: Meteorologia
Title: “Brazilian Intraseasonal Variability: Rainfall Patterns and Predictive Skill”
Advisor: Prof. Dr. Pedro Leite da Silva Dias - IAG/USP
Judging Committee:
- Prof. Dr. Pedro Leite da Silva Dias - IAG/USP
- Profa. Dra. Leila Maria Vespoli de Carvalho – UCSB - por videoconferência
- Profa. Dra. Michelle Simões Reboita - UNIFEI - por videoconferência
- Prof. Dr. Everaldo Barreiros de Souza - UFPA - por videoconferência
- Prof. Dr. Marcelo Barreiro Parrillo - UdelaR - por videoconferência
Abstract
This thesis presents an in-depth analysis of intraseasonal variability over Brazil, its influence on extreme precipitation events, and the predictability of rainfall at sub-seasonal to seasonal (S2S) timescale in socio-economically strategic regions. Due to Brazil's vast geographical extent coupled with its heterogeneous precipitation regimes and the lack of a widespread surface rain gauge network, remote sensing datasets were critically assessed to determine the most suitable database for capturing intraseasonal oscillations. Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (IMERG) was selected due to its high spatial resolution and near real-time updates, enabling its research and operational applications. To assess intraseasonal oscillation unique aspects over South America, four univariate regional indices of intraseasonal variability are developed using IMERG rainfall and Outgoing Longwave Radiation (OLR) data, filtered within the 30-90 day intraseasonal band. The indices were constructed using the traditional linear technique of Empirical Orthogonal Functions (EOF) and the unsupervised machine-learning Self-Organizing Maps (SOM) technique. The regional indices provide a more refined representation of the intraseasonal oscillations impacts on extreme weather events over South America compared to global Madden-Julian Oscillation (MJO) indices, such as the Real-Time Multivariate MJO (RMM) index. The regional indices effectively capture tropical-tropical and tropical-extratropical teleconnections in the form of Rossby wave trains, which modulate intraseasonal precipitation in South America. Furthermore, the nonlinear propagation of intraseasonal oscillations is better represented by the indices derived from the SOM technique. We also evaluate the performance of the state-of-the-art European Centre for Medium-Range Weather Forecasts (ECMWF) model in forecasting rainfall over selected regions in Brazil at S2S timescale and also the forecast skill of the regional intraseasonal oscillation index. The research highlights that the ECMWF model demonstrates strong forecasting skills in the northern coast of Northeast Brazil for up to four weeks, largely due to the accurate representation of tropical-tropical teleconnections, such as those associated with the MJO. However, in subtropical and extratropical regions, such as the South Atlantic Convergence Zone region and the Southern Brazil, the model's forecast skill decreases significantly after two weeks, largely due to challenges in simulating synoptic-scale systems and tropical-extratropical interactions. The predictability of the regional intraseasonal oscillation index is particularly robust for up to 10 days, offering critical insights for planning and decision-making regarding extreme weather events. This level of predictability could not be equally achieved using global indices like the RMM. The results highlight that the regional indices of intraseasonal variability proposed in this thesis can be applied operationally by forecasting centers in Brazil, allowing for more effective monitoring of the evolution of intraseasonal oscillation and its potential impacts on extreme events.
Keywords: Intraseasonal variability, Sub-seasonal to seasonal forecasts over Brazil, Extreme precipitation events, Regional precipitation indices