Torrential rains and flash floods affected Andalusia and Murcia on Friday 28th September. The rain came after an extensive period of drought. The press talks about 600 people evacuated and six casualties.
Photos on the press showed people evacuated with boats and mud and water covering streets and the ground floors of buildings:
How did ECMWF forecasts do?
The Extreme Forecast Index (EFI) showed high values three days in advance. The following pdf shows on the left panel the EFI (shading). The red colour indicates the areas where the EFI has its maximum value (EFI=1). The model climate on the right panel tells us the amount of precipitation for the 99th percentile of the climate, which in the area of the flood is between 10 and 20 mm/24h (green shading):
The EPSgram for a location on the Spanish South-east coast indicates (left panel below) that the accumulated precipitation for a 24 hour period are for the majority of the Ensemble Members well above the climate (thin green line). In fact the forecast for Friday 28th, shows that from the 25th percentile upwards, the accumulated amounts are above the 99th percentile of the climate:
Ensemble Cumulative Distribution Functions
The file on the right shows the EPS distribution for the same location as the EPSgram. The top panel is the feature we are interested in and it shows the Cumulative Distribution Function (CDF) for the 10 forecast runs all verifying on the 28-29th (accumulated) September. The climate is in black. All the forecast runs indicated that there would be more rain than it would be expected if we took the climate as forecast. A particular shift to the right happens for the d+3 forecast range with a bunch of curves that move to the right suggesting very large extreme values (>50mm/24h). There is some spread in the EPS members as to what it should be the amounts of precipitation: for example following the red curve one notices values from >10mm/24h to >60mm/24h: they are obviously associated to a different probability of happening!