Would like to share on two episodes of monsoon surges over Singapore in January this year. The first monsoon surge had a highest daily total rainfall of 238.4mm and 210.6mm on 1 and 2 January. The second monsoon surge, which occurred between 8 and 13 January, saw a highest daily total rainfall of 204.0mm on 10 January. The surges were quite severe. Within the first two weeks of January, the total rainfall recorded at our Climate Station reached 648.4mm, making this the wettest January in the last 30 years.  

The ECMWF model runs had quite a big forecast skill difference for both surge episodes. The meteogram for the first surge showed a large spread with the outliers reaching up to 16-18mm range and a median of around 3mm for the precipitation. The meteogram for the second surge, however, had a larger spread with the outliers reaching up to 60-80mm range and a higher median. Deterministic runs forecast the second surge to be more intense and longer lasting. Wind patterns were quite similar for both surge episodes, with deep troughing and sharp cyclonic shear at the low levels and no significant upper level easterlies. 

Despite models indicating that the second surge would be more severe, more rainfall actually fell over Singapore during the first surge. It was also observed that the precipitation eased off on some days during the second surge episode.

A few possible reasons for the models behaviour were identified. Firstly, the MJO phase was weak for the first surge episode, but entered into the active phase over the Maritime continent for the second surge episode. Secondly, the sudden stratospheric warming (SSW) over the Artic that started around 5 January this year might have contributed to a stronger cold air outbreak over the Maritime Continent and led to more intense precipitation signals over Singapore and the surrounding region.

Below are some questions for discussion.
1) Is SSW is a contributor? How can we verify it?
2) Are there any other reasons for such big forecast skill difference?
3) Why did the first surge episode turn out to be a stronger event?
4) The deterministic runs forecast the intense spell to last from 8 to 13 January. However, the rainfall that fell over on most days were relatively light. What are some better ways to forecast between continuous rain and isolated passing showers with enough lead time?

Welcome any ideas. Thank you.



3 Comments

  1. Rebecca Emerton

    Thank you for this really interesting post and overview of the events - we have not previously looked at these events in particular, and are looking into this through the Severe Event Catalogue. A link will be added soon once we have more information. One of our research scientists suggests that for your question (1), a way to verify if SSW is a contributor could be to consider the ensemble members that had a good SSW prediction, compared to the ensemble members that had a good SSW prediction, and evaluate the difference in precipitation forecasts for these two sets of ensemble members. 

    Do you have any additional insights or comparisons with other global or regional models - it would be interesting to help understand the reasons for this, if there is any information on whether other models had similar issues for these two monsoon surges. 

  2. Linus Magnusson

    Hi, Thanks a lot for your post! As a start, I have made two severe event pages for the two episodes: 

    202101 - Rainfall - Singapore Part 1

    202101 - Rainfall - Singapore Part 2

    The higher predictability for the second event comes out well, but for neither of the events the ensemble mean breaks the 99th percentile of the model climate for 3-day precipitation. However, for the second event some really extreme members existed, as seen in the forecast evolution plot of the Part 2 page.

    Looking at the HRES maps of 3-day precipitation, I noticed a strong land-sea contrast. Is that seen as a problem in the ECMWF model for this region? How does it compares with your regional models?

    Best Regards,

    Linus

  3. NEA MSS

    Dear Rebecca, our regional models did pick up active weather over the sea areas, but they appeared as passing showers over land due to the strong winds for the first surge episode. Continuous rain lasting for several hours was forecast for the second surge episode. In terms of the severity, there’s some similarities between EC global model and our regional model.