Status:Finalised Material from: Linus
1. Impact
On 19-20 July 2016 a severe rainfall hit China.Preliminary figures says 14 people dead and 7100 houses affected but landslides. The precipitation in the Beijing metropolitan reached 300 mm.
http://www.shanghaidaily.com/nation/14-dead-72-missing-in-north-Chinas-heavy-downpour/shdaily.shtml
http://www.bbc.co.uk/news/world-asia-china-36852886
2. Description of the event
24hr precipitation map from 19/07 at 05:00 to 05/20/07 at 05:00 (UTC+8)
The plots below show the evolution of MSLP and 12-hour precipitation from 18 July 00z to 20 July 12z
The plots below show the same as above but for z500 and t850.
3. Predictability
3.1 Data assimilation
3.2 HRES
The plots below show the 24-hour precipitation (19 July 12z-20 July 12z) and MSLP for 20 July 00z in HRES forecast from different initial times (see title in figure). Te first panel shows the precipitation observations. The forecasts were very consistent from 16 July 00z and onwards. Before that the forecast had the cyclone on a southerly path.
3.3 ENS
The plots below shows the EFI and SOT for 3-day precipitation 19-21 May. Between 15 and 16 the area of EFI shifted northward to include Beijing area.
The next plot shows the evolution of the ensemble and HRES for 48-hour precipitation over Beijing. The ensemble started to pick up the signal of extreme rainfall from Saturday 16 July 00z with 25% risk for more than 150 mm.
3.4 Monthly forecasts
The plot below show weekly precipitation anomaly for 18-24 July. The forecast from 14 July (second panel) had the anomalous precipitation further south. Also note the early prediction of the dry anomaly further south.
3.5 Comparison with other centres
In the plot below forecasts from the TIGGE archive is compared for the 48-hour rainfall for Beijing from different centres. The plot includes ECMWF (red), UKMO (blue), NCEP (green) and CMA (purple). As discussed above, ECMWF got a strong signal for the rainfall from 16 July (Saturday). UKMO picked up the signal around the same time, but clearly underestimated the rainfall in later forecasts. NCEP picked up the signal 1.5 day later (Monday) and similar for CMA.
4. Experience from general performance/other cases
- Compare with case from 21 July 2012, discussed in http://dx.doi.org/10.3402/tellusa.v68.31045
5. Good and bad aspects of the forecasts for the event
- Good prediction of the rainfall from 16 July
- Earlier forecasts had the precipitation further south