Material from: Linus, Esti


 


1. Overview

Between 30 September and 4 October 2026, Crete was hit by five consecutive days of intense rain and thunderstorms. Authorities placed the whole island on red alert, schools in Chania and Rethymno were closed on 1 October, and repeated 112 emergency alerts were recorded in parts of Rethymno, Heraklion and Lasithi. The highest totals fell over the mountains: more than 220 mm was reported at Profitis Ilias (Heraklion) on 1 October, and the Lasithi Plateau received about 440 mm in two days and around 470 mm in four days, compared with a normal annual total of 700-900 mm. Damage was reported in all four regional units. In Mylopotamos (Rethymno), the village of Zoniana was flooded, the Diskouri bridge near Livadia was swept away and 30-40 residents were evacuated. An 80-year-old livestock farmer was found dead near Karines in Rethymno after being swept away by a stream. In Lasithi, landslides and torrents damaged roads and tourist areas around Ierapetra, and up to 85% of the crops on the Lasithi Plateau may have been affected. Between 1 and 4 October the Fire Service received 485 calls and moved 101 people to safety.

Sources: Euronews, ProtoThema, Greek City Times

2. Analyses and Observations

2.1 Event Definition

The evaluation will focus on the 5-day precipitation between 30 September 06UTC and 5 October 06UTC for a 0.5x0.5 degree box centred on 35.3N, 24.4E on Crete.

2.2 Analyses

The plots below show analyses of MSLP and 6 hour rainfall from 28 September 00UTC to 5 October 00UTC, every 12th hour.

The plots below show z500 and T850 from 24 September to 5 October 00UTC, every 24 hour. During the episode a trough build in from north-west on 26 September with a second centre on 30 September that stayed stationary until 5 October.

2.3 Observations


2.4 Climatological perspective

3. Forecasts

3.1 Data assimilation and observation usage

 

3.2 Single Forecasts for the event based on 00UTC forecasts (defined in Section 2.1)

Observations and analysis for the event

Looking carefully it seems like the ECMWF analysis and all forecasts below placed the most intense rainfall somewhat to the north of the maximum observations.

Control forecast (IFS 9-km resolution)

DestinE (IFS 4.4km resolution)

AIFS-single (AIFSv1.0 ~0.25 resolution)

AIFS-CRPS ensemble (~0.25 degree resolution)

3.3 Ensemble distribution

EFI (based on IFS 9-km ensemble)

The plots below show the EFI for 5-day precipitation for 30 September 00UTC to 5 October 00UTC, from different initial dates. There was an early signal about a wet period in the eastern Mediterranean but the strong AFI signal for Crete appeared around 27 September.

Forecast Evolution plot

The plot below shows the evolution of the forecasts for 5-day precipitation (30 September 06UTC - 5 October 06UTC) for the box on Crete defined in Section 2.1. The signal grew gradually from 25 October and onwards than it seems like IFS was somewhat ahead of AIFS ensemble. One can note that the model climate maximum is very extreme for this event definition.

3.4 Sub-seasonal forecasts

 The figures below shows verification plots for weekly mean precipitation valid 28 September to 5 October. The strong signal appeared in the forecast from 25 September, but also week 2-3 forecasts indicated a wetter than normal period in central-eastern Mediterranean.

3.5 Comparison with other centres


4. Hazard Forecasts


5. Dedicated Experiments


6. Event Summary

6.1 Good and bad aspects of the forecasts for the event

  • IFS ensemble gave an early signal with the EFI for the 5-day precipitation over Crete was 0.6-0.8 in the forecast from 26 September 00UTC (4 days before the start of the event) 
  • The short-range IFS forecast (HRES from 30 September 00UTC, 213 mm) matched the box observations well. However, all models put the maximum along the north coast and over the sea north of western and central Crete, and underestimated the rainfall over eastern Crete (Heraklion-Lasithi), where stations reported 200-400 mm while the forecasts gave mostly 50-150 mm. Potential overestimation of the rainfall along the coast, for example Chania.
  • IFS seemed to pick up the signal earlier than AIFS and also had a more rainfall totals in general.
  • Some sub-seasonal signal for a wet period 2-3 weeks before the event.

6.2 Experience from general performance/other cases


7. Additional material