Extreme rainfall events over Rio de Janeiro State, Brazil: Characterization using probability distribution functions and clustering analysis

• Published in: Atmospheric research Vol. 247; p. 105221
• Main Authors: Lima, Allana Oliveira, Lyra, Gustavo Bastos, Abreu, Marcel Carvalho, Oliveira-Júnior, José Francisco, Zeri, Marcelo, Cunha-Zeri, Gisleine
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-01-2021
• Subjects: Climate extremes; Cluster analysis; Cumulative distribution function; Daily rainfall; Recurrence interval; Climate extremes; Cluster analysis; Recurrence interval; Daily rainfall; Cumulative distribution function
• ISSN: 0169-8095; 1873-2895
• DOI: 10.1016/j.atmosres.2020.105221

Extreme rainfall events are likely to become more frequent according to recent scenarios of climate change. This issue is especially important over regions with complex topography, which enhances rainfall variability when associated with weather patterns. The state of Rio de Janeiro (SRJ), southeastern Brazil, is characterized by altitudes ranging from the mean sea level up to 2500 m.a.s.l, in mountain ranges and valleys covering significant parts of the region. Time series data of annual maximum daily rainfall were obtained from 110 stations with a data coverage of at least 20 years, from 1960 to 2010. The Probability Distribution Functions (PDFs) normal, log-normal, exponential, gamma, Gumbel, Weibull, and Generalized Extreme Value (GEV) were fitted to maximum rainfall series. Goodness-of-fit tests (Chi-squared - χ2 and Anderson-Darling) revealed that the Gumbel, GEV, and log-normal were found to be the best choices. However,the Gumbel and GEV PDFs were the best ranking by the χ2 and Anderson-Darling test, respectively. Extreme rainfall events with different recurrence intervals (5, 10, 25, 50 and 100 years) were calculated based on the Gumbel and GEV Cumulative Distribution Function (CDF). The differences between extreme values from Gumbell and GEV function increased as the shape parameter increases from zero, with higher probability and extreme value. Five regions with homogeneous patterns of extreme rainfall were identified using clustering analysis (Ward's method) and different recurrence intervals. Overall, the regions with higher values of extreme rainfall in all scenarios and CDFs were the ones close to the coast, within 40 km, and south of Serra dos Órgãos mountain range, located in the middle of the state. The mountain range separates the state in two halves, concentrating higher values of extreme rainfall in the lower part, where the city of Rio de Janeiro, the state's capital, is located. Scenarios for both CDF (GEV and Gumbel) indicated daily rainfall events up to 200 mm, with recurrence intervals of 50 to 100 years. In addition, the southernmost part of the state is subjected to rainfall extremes up to 260 mm in scenarios of 50 to 100 years of recurrence interval. This region, and the state's capital, are characterized by complex topography and a high fraction of population living in slums over hills, or lowlands near the ocean, increasing the vulnerability to events such as landslides and floods associated with extreme rainfall.
[Display omitted]
•Seven probability distribution functions were fitted to annual maximum daily rainfall series in Rio de Janeiro State, Brazil•The best-fit distributions (Gumbel and GEV) were used to estimate extreme rainfall with different recurrence intervals•The cluster analysis and spline tension method were applied to assess the spatial patterns of the extreme rainfall•The regions with higher extreme rainfall were close to the coast and south of mountains located in the middle of the State•In these regions, extreme daily rainfall range between 200 and 300 mm, with return times of 50 to 100 years were expected