A carregar...
Publicação
Análise de temperaturas extremas em Lisboa
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 7.74 MB | Adobe PDF |
Orientador(es)
Resumo(s)
As temperaturas extremas são acontecimentos que influenciam significativamente as po-
pulações nas mais diversas áreas, que no contexto presente das alterações climáticas, só
tenderão a aumentar em frequência e intensidade. Assim como toda a região do Mediter-
râneo, Lisboa também é um local suscetível a um aumento de ocorrências desta natureza.
Os modelos clássicos da estatística podem, no entanto, subestimar a probabilidade de
eventos extremos, uma vez que se concentram sobretudo no comportamento médio dos
dados. A Teoria de Valores Extremos surge então como uma alternativa.
Neste trabalho modelaram-se os dados da série diária das temperaturas máximas
de Lisboa para o período de 1856–2019, através de metodologias da Teoria de Valores
Extremos, nomeadamente, o método dos blocos, com a distribuição de Valores Extremos
Generalizada (GEV), e o método peaks-over-threshold (POT), com a distribuição de Pareto
Generalizada (GP). Os estimadores dos parâmetros foram obtidos através do método de
máxima verosimilhança. Além dos modelos estacionários, foram também considerados
cenários de não estacionariedade, com modelos que incorporam tendências ao longo
do tempo. Os resultados obtidos neste trabalho indicam que os modelos GEV e GP,
tanto estacionários quanto não estacionários, descrevem adequadamente as temperaturas
extremas. Destes, os modelos não estacionários representaram uma melhoria face aos
estacionários na modelação dos dados. Adicionalmente, foram calculados níveis de retorno
com períodos de retorno de 10, 20, 50 e 100 anos.
Extreme temperatures are events that significantly affect populations across multiple sectors and, in the current context of climate change, are expected to increase in both frequency and intensity. As with the entire Mediterranean region, Lisbon is also highly susceptible to such occurrences. Classical statistical models, however, may underestimate the probability of extreme events, as they primarily focus on the central behavior of the data. Extreme Value Theory thus emerges as a suitable alternative. In this work, the daily maximum temperature series for Lisbon, covering the period 1856–2019, was modeled using Extreme Value Theory methodologies, namely the block maxima method with the Generalized Extreme Value (GEV) distribution and the peaks- over-threshold (POT) method with the Generalized Pareto (GP) distribution. Parameter estimation was carried out via the maximum likelihood method. In addition to stationary models, non-stationary scenarios were also considered, incorporating trends over time. The results obtained in this work indicate that both stationary and non-stationary GEV and GP models adequately describe temperature extremes, with the non-stationary models providing a better fit. Furthermore, return levels were estimated for the 10, 20, 50, and 100 years return periods.
Extreme temperatures are events that significantly affect populations across multiple sectors and, in the current context of climate change, are expected to increase in both frequency and intensity. As with the entire Mediterranean region, Lisbon is also highly susceptible to such occurrences. Classical statistical models, however, may underestimate the probability of extreme events, as they primarily focus on the central behavior of the data. Extreme Value Theory thus emerges as a suitable alternative. In this work, the daily maximum temperature series for Lisbon, covering the period 1856–2019, was modeled using Extreme Value Theory methodologies, namely the block maxima method with the Generalized Extreme Value (GEV) distribution and the peaks- over-threshold (POT) method with the Generalized Pareto (GP) distribution. Parameter estimation was carried out via the maximum likelihood method. In addition to stationary models, non-stationary scenarios were also considered, incorporating trends over time. The results obtained in this work indicate that both stationary and non-stationary GEV and GP models adequately describe temperature extremes, with the non-stationary models providing a better fit. Furthermore, return levels were estimated for the 10, 20, 50, and 100 years return periods.
Descrição
Palavras-chave
Temperaturas extremas Teoria de Valores Extremos Método dos blocos GEV Peaks-over-threshold GP