Status:Ongoing analysis Material from: Linus, Fernando, Tim H.
1. Impact
The SE Brazil, in particular the Rio Grande do Sul state, was badly hit by torrential rains in the last days of April and the beginning of May, with more than 500mm/5 days reported. The devastation is impressive; towns under water and thousands of people without power or communication and displaced. There has been a lot of damage and loss of life in a number of areas. One was the small town Roca Sales (picture below), where overflowing river water appears to have demolished houses. So far as I can tell this is not related to the partial dam break reported in the media.
2. Description of the event
For the rainfall evaluation we will focus on the 3-day period 30 April-2 May for 0.5x0.5 degree boxes around Caxias (29.2S, 51.2W) and Santa Maria (29.7S,53.7W).
The plots below show analyses of MSLP and 6 hour rainfall from 28 April 00UTC to 3 May 00UTC, every 12th hour. Box box outlined represents Caxias. Note that the heavy rainfall started already on 28 April.
The plots below show analyses of T850 and z500 from 27 April to 3 May.
3. Predictability
3.1 Data assimilation
3.2 HRES
The plots below show 72-hour precipitation (30 April 00UTC - 3 May 00UTC) in observations (first plot) concatenated short forecasts (second plot) and ENS control forecasts with different lead times. The box marks 0.25x0.25 degree centred on Caxias. The analysis plot based on 6-hour forecasts clearly underestimated the rainfall around Caxias, while the forecasts more or less missed the extreme on all lead times but instead had the maximum rainfall further to the west.
The plots below shows the same as above but forecasts from DestinE. The performance for Destine (4.4km) is similar to the 9-km version above.
The plots below shows the same as above but forecasts from AIFS. These forecasts also had the maximum rainfall shifted to the south, but the smoothness might have helped to bring up the totals around Caxias.
3.3 ENS
The plots below shows EFI for 3-day precipitation (30 April-2 May) from different initial dates. There was an early signal in the EFI, but in the short-range the rainfall shifted to the south and did not saturate towards 1.
The plots below shows EFI for 3-day integrated water vapour flux (30 April-2 May) from different initial dates.
The plot below shows the forecast evolution plot for 72-hour precipitation for Caxias.
Observation - green hourglass
Analysis - green dot
ENS control–red
DestinE4.4km - purple dot
ENS - blue box-and-whisker
AIFS - cyan dot
Model climate – cyan box-and-whisker
Ensemble mean as black diamonds. Triangle marks the maximum in the model climate based on 1800 forecasts.
While predicting extreme (around 99th percentile of the model climate) for Caxias, the forecast did not predict anything exceptional as seen in the observations. One can note that AIFS was somewhat higher than the ensemble mean for most forecasts.
The plot below shows the same as above but for Santa Maria. The plot from the concatenated analysis included to show the location of Santa Maria.
For Santa Maria the extreme was somewhat better captured, with the shortest forecast having an ensemble mean above the maximum in the model climatology but still only half of the measured 3-day precipitation. One can note that the control forecasts was always on the lower end of the ensemble in the short-range forecasts.
3.4 Monthly forecasts
The plots below show forecasts of weekly precipitation anomalies for 29 - 5 May. Already the longest forecast (5-week) predicted a wet anomaly in the region.
The plots below show forecasts of weekly z500 anomalies for 29 - 5 May.
3.5 Comparison with other centres
4. Experience from general performance/other cases
5. Good and bad aspects of the forecasts for the event
- Also the concatenated 6-hour forecasts missed the northerly extent towards Caxias
- The magnitude of the event was underestimated in the short- and medium-range.
- Early signal for wet period in extended-range.