|
Descripción
|
|
|---|---|
| A physically based nonstationary flood frequency analysis using the Gumbel distribution is proposed. The modelling is based on the primary drivers that generated the long-term oscillations of flood time series. A Generalized Linear Model is used to predict the location parameter of the Gumbel probability distribution. The definition of the explanatory variables is performed using the Bayesian Information Criteria. The results indicate that the antecedent catchment wetness condition preceding the occurrence of flood events is the most critical driver of long-term flood changes in the study area. Additionally, some climatic features and short-term precipitation are relevant particularly in regions where the influence of atmospheric systems originated on the Amazon forest are most noticeable. The results also show that both stationary Gumbel models overestimate floods in the region and the non-stationary approach gives a more realistic picture of the flood risk. The non-stationary model supplies narrow prediction intervals and reproduces accurately long-term oscillations. The prediction results obtained in this study can be useful in the operation of flood forecasting systems, as it can anticipate from two to seven months the flood occurrence with reasonable accuracy. | |
|
Internacional
|
No |
|
Nombre congreso
|
I Encontro Nacional de Desastres |
|
Tipo de participación
|
960 |
|
Lugar del congreso
|
Porto Alegre, Brasil |
|
Revisores
|
Si |
|
ISBN o ISSN
|
|
|
DOI
|
|
|
Fecha inicio congreso
|
25/07/2018 |
|
Fecha fin congreso
|
27/07/2018 |
|
Desde la página
|
1 |
|
Hasta la página
|
8 |
|
Título de las actas
|
I Encontro Nacional de Desastres. Proceedings |