Keywords
Flash flood; LSTM; Uncertainty; Precipitation nowcasting
Start Date
5-7-2022 12:00 PM
End Date
8-7-2022 9:59 AM
Abstract
Flash floods are among the most prompt and destructive natural phenomena. To issue warnings on time, various attempts were made to extend the forecast horizon of flash floods prediction models. Particularly, introducing rainfall forecast into process-based hydrological models was found effective. However, integrating precipitation predictions into data-driven models has not been fully addressed yet. In this paper, we assess the effectiveness of introducing future rainfall measurements into a data-driven model while taking forecast uncertainties into account. Computational experiments showed that introducing future rainfall observation (as perfect prediction) significantly improved the predictive performance of the flash flood model for extended lead times. We also found that errors in rainfall forecast measurements had a larger impact on the predictive ability of the proposed model as opposed to errors in the time of occurrence
Integrating precipitation nowcasting in data-driven prediction of extreme hydrological events
Flash floods are among the most prompt and destructive natural phenomena. To issue warnings on time, various attempts were made to extend the forecast horizon of flash floods prediction models. Particularly, introducing rainfall forecast into process-based hydrological models was found effective. However, integrating precipitation predictions into data-driven models has not been fully addressed yet. In this paper, we assess the effectiveness of introducing future rainfall measurements into a data-driven model while taking forecast uncertainties into account. Computational experiments showed that introducing future rainfall observation (as perfect prediction) significantly improved the predictive performance of the flash flood model for extended lead times. We also found that errors in rainfall forecast measurements had a larger impact on the predictive ability of the proposed model as opposed to errors in the time of occurrence
Stream and Session
false