Monday, November 28, 2011

Fishermen died unnecessarily in Tropical Cyclone 5

In my post of a few days ago I wrote about a developing tropical cyclone near Sri Lanka. This storm has since developed and been named "Tropical Cyclone Five" by the Joint Typhoon Warning Center. It's not particularly intense as tropical cyclones go; if it were in the Atlantic we would say it was a "Tropical Storm". Still, that's enough to be dangerous if you sail a boat into it. Unfortunately, a number of Sri Lankan fishermen have died - exactly how many is not clear yet, but the media are reporting 20 deaths and 43 missing - apparently because they did just that. Here is a recent story about it.

Had these fishermen known there was a serious storm there, presumably they might have stayed home. Almost certainly they did not know, because no forecast of this system was issued in Sri Lanka. Apparently Sri Lankans realize that they deserve better; a government inquiry has been launched into why their Department of Meteorology didn't predict TC 5.

Sometimes tropical cyclone formation is very difficult to predict. But as science improves year by year, we get better at it as our tools for doing it improve. The real tragedy of TC 5 is that it seems to have been unusually predictable, given the tools we - the international scientific community -currently have and were actually using in the lead-up to the storm's development.

We here in the Maldives working on the DYNAMO project are continuously looking at numerical model forecasts for the region. Some of these forecasts are being done specifically for the project, but most are the same ones that are done all the time by various operational weather centers around the world. They are being made a little more directly and easily accessible to us than they normally might be, but much of this information is routinely available on the web.

We saw TC 5 forming in some of the model forecasts around a week ahead of time, more or less simultaneous with the predicted onset of the active MJO phase in the Indian Ocean. This is a long way ahead to predict tropical cyclogenesis, so at first we weren't confident the models were right. However, as the week progressed, the picture became more and more convincing. The different models were agreeing with each other, and forecasts from the same model on different days also were very consistent.

Here is the forecast of relative humidity and wind at 850 hPa (around 1.5 km above the surface) for 06 UTC on November 25th. This is the GFS model, run by the US National Oceanic and Atmospheric Administration. The plot shown is made from a model run six days earlier. That is, it's a six-day forecast. Notice the swirling arrows right on the southern tip of India. That's the predicted TC 5. Sri Lanka is just to the southeast of the tip of India, well within the forecast storm.
Now here's a set of five more predictions, all valid at the same time (06 UTC on the 25th), but performed on successive days, getting closer and closer to the 25th. So the first one down is a five-day forecast, the next is a four-day forecast, etc. Keep your eyes on the tip of India - around 10N, 75E.
You can see the consistency in the forecasts. The predicted strength of the storm increased a bit in the later forecasts compared to the six-day one, but its position and timing didn't change much.

It seems to me that if forecasters in Sri Lanka had been looking closely at this model --- or others like it that showed more or less the same thing --- they couldn't have avoided thinking that there might be something here worth telling the public about. Even if one were disinclined to trust the models, the consistency of their message should draw one's attention. So I can only guess that they weren't looking at the models.

I was communicating by email with a colleague in Sri Lanka earlier in the week. On Tuesday I mentioned that the models were showing a TC developing, and on Wednesday the 23rd - as it looked more and more real - I asked him whether anyone there was paying attention to it (having looked on the web to see if any forecast had been issued, and seeing none). He called some contacts at the Department of Meteorology, and said that "they are expecting heavy rains on the eastern hill slopes but not a cyclone or storm - looks like they are only looking at the satellite images". In many cases tropical cyclones form far from shore and so one has time to see them develop in satellite images well before they form a threat to those on (or near) land. TC 5 formed close to shore so this method didn't work well - one needed the models.

Well, we - my Sri Lankan colleague and I - tried, in a very small way. In hindsight, maybe we could have tried to find some way to make a louder warning to someone who might have been able to do something. But neither of us is a forecaster by trade nor in any position of authority. Speaking for myself, I just couldn't quite believe it could be true that I had information that could save lives, but that no one else who knew as well or better was minding the store. Apparently that was exactly the case though. It wasn't my job, and I'm pretty sure I couldn't have done anything for those fishermen, but I can't help feeling that I could have.

1 comment:

  1. That's such a shame that the forecast office messed this up.