Assessing the uncertainties related to the temporal evolution of the rainfall velocity field in radar nowcastings

Flooding is a very common natural disaster around the world, putting local population at risk and heavily affecting the economy. Due to climate change, its frequency and intensity is expected to increase. In urban areas, the impact of heavy rainfall is even higher, once there is a greater area of impermeable surfaces and the large concentration of people. Being so, to accurately predict extreme rainfall events hours ahead is very important, permitting to effectively respond to emergency situations. However, rainfall forecasting are subject to different sources of uncertainties.

Nowcasting models are short-term rainfall forecasts capable of producing predictions a few hours ahead. It has a central role to predict flash floods in urban areas due to its high spatial and temporal resolution. Nonetheless, nowcasts are subject to uncertainties due to radar errors, temporal development of velocity fields and temporal evolution of rainfall. Nowcasting models assume that the temporal development of velocity fields are stationary. This assumption contribute to uncertainties in the forecast after 1h lead-time, with a much bigger impact after 3h lead-time. In an attempt to deal with this source of errors, probabilistic forecasts are produced by randomly adding noise to the forecast.

The aim of this work is address the uncertainties related to the temporal development of rainfall velocity fields, adding noise in realistic way to generate ensembles. Radar data from the UK Met Office, available at the British Atmospheric Data Centre (BADC) website was used. The radar data has a temporal resolution of 5 min and a spatial resolution of 1 km x 1km. Radar data from 20 events with stratiform and convective rainfall from 2008 were selected. The rainfall velocity fields were calculated through the nowcasting model. Rainfall forecasts are generated using motion fields from the last few forecasts to produce ensembles that take into account the uncertainty in the motion fields. Rainfall velocity fields are calculated using radar scans that are 5 min, 10 min and 15 min apart from each other. Each velocity field can produce a new ensemble. Preliminary results show that there is a strong influence of the rainfall velocity fields in the forecast with a few hours of lead-time. This poster presents the results of this new method to generate ensembles assessing their value compared to deterministic rainfall forecasts.

The first named author is grateful to CAPES – Ciencia sem Fronteiras for funding this PhD research.