Detection and attribution of changes in small and large floods across Brazil

Código

I-EBHE0154

Autores

Gabriel Anzolin, Vinícius Bogo Portal Chagas, Pedro Luiz Borges Chaffe

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

Understanding changes and generation mechanisms of floods is key to better water resources planning and disaster preparedness. Typically, the detection and attribution of flood change focus on the average flood behavior, without distinguishing between small and large floods. Here, we investigate: (i) changes; and (ii) the role of flood generation mechanisms such as event rainfall peak and antecedent wetness in explaining annual maximum flood (AMF) variability using 737 hydrographic basins from Brazil for 1980-2018 period. We detect and attribute flood change throughout linear quantile regression. Two quantiles of interest are selected: the 0.50 quantile, representing the median flood behavior or small floods; and the 0.98 quantile, representing large floods. For the detection of change, we assume a log-linear dependence between floods and time; for the attribution analysis, we assume log-log dependence between floods and event rainfall peak and antecedent wetness simultaneously. Our results show that both small and large floods have been changing across Brazil. Small and large floods have similar magnitudes but opposite signs in Western NE Atlantic, Parnaiba, Southern Brazil (SB), and Southern Amazon (SA). In the basins of Sea Ridge (SR), large floods increased more than small floods. Overall, event rainfall and antecedent wetness drive flood change in SB and SR for both small and large floods. In most of the NE, small floods are modulated by decreasing antecedent wetness. Despite the increases in large rainfall peaks over the entire NE, large floods are increasing only in basins from Western NE Atlantic and Parnaiba regions, while large floods in the remaining basins are also decreasing due to large decreases in antecedent wetness. Our findings suggest that inferring large flood change and its generation mechanisms from the average flood behavior can be misleading. We argue that distinguishing between small and large floods should be considered to support flood prediction and climate-change impact studies.