Status:Ongoing analysis Material from: Esti
1. Impact
Last week, Sri Lanka faced very severe floods following heavy rains since Sunday, 2 June. Over 159,000 people were affected cross 13 districts. Colombo, Gampaha, Kalutara, Matara, Ratnapura, Kegalle, and Galle are the worst hit areas.The floods have caused at least 17 deaths and five people to be missing.
2. Description of the event
The monsoon rains began 3 weeks ago in that area, but intensified over that weekend, leading to a record 400 millimetres of rain in parts of the country which has led to floods and landslides. Flood alerts were in place for Colombo and other areas (although I could not see anything in Glofas, probably because they are small catchments). This was a major and important event in the country. . It seems that the registered precipitation was quite high on the 1st June, but also on the 2nd. Maps of the accumulated precipitation on the 1st June show peaks of more than 250 mm in areas of the southwest of Sri Lanka, and more than 200 mm the following day. In the Ogimet webpage I also found some available stations, but the maximum amount on the 1st is 163 mm an 52 mm on the 2nd June.
3. Predictability
Here we can find the evolution plot for 24h accumulated precipitation in the location were the maximum precipitation is registered in the ENS system valid on the 1st June, and another evolution plot for the mean of 24h precipitation in the area were the maximum precipitation was observed (in the map plot for the 1st June, the area with dark purple, purple and dark red, which correspond to values greater than 100 mm). The results are pretty similar in terms of predictability: some ensemble members (but just the tail of the distribution) were providing good guidance of some extreme precipitation event even 10 days in advance, around the 21st June, but it was not until the 26th June when some other ensemble members also indicated larger values. Then, from the 30th June 00 UTC forecast, all the ensemble members decreased the precipitation considerably, which is quite remarkable in the mean area plot. It is also clear in the specific location of maximum precipitation in the ENS, however in the last forecast run before the event, the ensemble again increased the precipitation, so it was a kind of jumpiness in the forecast. HRES forecast (red dots) behaved in a similar way. Progressively increasing the precipitation until 29th June 00 UTC run, then decreasing it again. This jumpiness is also observed in the DestinE forecast.
3.1 Data assimilation
3.2 HRES
The underestimation of the precipitation amounts is pretty clear by both forecasting models at different lead times (T+24h, T+48h and T+96h). We could say that, in general, DestinE provides slightly better guidance as it increases the maximum precipitation in the most affected areas, but this does not happen in all the lead times (at T+48h ,the precipitation is larger in HREs than in DestinE, for example)
3.3 ENS
3.4 Monthly forecasts
The signal of much wetter than normal in that specific region of Sri Lanka is very well defined and highlighted in both maps, in week 1 and week 2. I didn't check week 3 but it may be worth to see how far the signal is still maintained event at week 3 for this very local event.
3.5 Comparison with other centres
4. Experience from general performance/other cases
5. Good and bad aspects of the forecasts for the event