Wednesday, November 17, 2010

Cyclone and weather forcasting

About 10 days ago we were warned that Cyclone Jal would hit North TN and South Andhra Coast. While the cyclone turned out to be a damp squib, what I found surprising is the extremely sloppy weather forecasting. As many people pointed out, the centre of the cyclone was already above Chennai sometime on Sunday 7th November afternoon. This was clear from the satellite picture. However the weather office kept predicting that the cyclone would cross the coast Sunday night. How could there be so much discrepancy? Or was the outline of the country wrongly superimposed on the cloud map.

Moreover consider the weather forecast. Here are a couple of samples.

Under its influence, rain/thundershower would occur at most places with heavy to very heavy falls at a few places and isolated extremely heavy falls (≥25 cm) over north Tamil Nadu and Puducherry during next 24 hours and at most places with isolated heavy to very heavy falls over Rayalaseema and South Coastal Andhra Pradesh during the same period.

Rain/thundershowers at most places with heavy to very heavy falls at a few places would occur over South Interior and Coastal Karnataka during next 48 hours. Isolated extremely heavy falls would also occur over south Interior Karnataka during next 24 hours


Apart from the completely arbitrary use of lower and upper case, it seems to me that the IMD has a simple algorithm of concatenation of phrases. These are of two kinds a) (light, moderate, heavy, very heavy, extremely heavy, isolated extremely heavy) rain and b) (few, some, many, most) places. Combine one phrase from a) and another from b) and you have got a prediction, even though in actual content it leaves the hapless residents in those places totally at sea (regrettably sometimes literally).

Surely Mausam Bhavan and our own Chennai weather forecasting stations under S. R. Ramanan are capable of somewhat more precise forecasting? A lot of it appears to be lazy thinking. If you say isolated rain in some areas, moderate in a few, heavy in isolated pockets, you have covered most possibilities and nobody can accuse you of having got it wrong. I think our weather men could try a little harder in this day and age.


Ludwig said...

Cyclones are somewhat of a pet subject, so let me vouchsafe unto you this quite useful resource: the Joint Typhoon Warning Center. It is a US Navy facility of some sort, and they provide great satellite pictures, a telegram-ish text forecast, and a rather nice graphic which shows the track of the storm, where it's at, and a "probabilistic cone" of where it might be headed. Do check out!

Rahul Basu said...

Ludwig: Thanks for the comment and link but as it happens I look at that site quite often, they are indeed as you say very informative and well presented -- in fact whenever there are reports of a cyclone developing in the Bay of Bengal expected to hit Tamil Nadu coast I tend to follow it here.

Sourendu said...

I keep ranting about the IMD web site, so it is a great pleasure to find you doing the same. All the data on this site seems to be entered by hand. There are graphs drawn by hand and then scanned and put on the site! What does it take to download free graphing software?

There is a sub-site which gives 50 year's rainfall data for various cities. Now it is very hard to cross check something like this, but the number of extreme weather events reported there can be checked against other compilations. When I did so, it turned out that the IMD archive did not agree with others.

If they can't even transcribe numbers properly, what can you believe them on?

William of Occam said...

Even more regrettable is the fact that in 2005, India's Dept. of S&T banned annual long range monsoon forecasting by any agency other than the Indian Meteorological Department.
I understand the motivation for this rule (prevent market instability) but saddened that a proud democratic nation needs to do this.

Anant said...

It is not just IMD. Most Government and quasi-Government agencies have atrocious web-sites. Try visiting some of them and finding anything useful. In the area of weather forecasting there are fantastic web-sites. Why can't our IMD simply outsource the web-site maintenance if they are too stupid to maintain a decent one themselves?

Rahul Basu said...

It seems I have opened the flood gates to a lot of complaints that people have about IMD. In fact, I recall that IMD was one of the first to get a Cray supercomputer in the late eighties when they were banned from being exported to India due to dual-use issues. They have, in other words, never been short on resources. It's therefore ridiculous that they should function in this lazy, unprofessional and bureaucratic manner.

Sourendu: Their graphs are nowadays not hand drawn any more - however they look like they were made on a plotter bought in the 80s or 90s. Given how advanced weather sites are nowadays, IMD has no excuse for such poor quality presentation and forecasting.

Anant said...

``...function in this lazy, unprofessional and bureaucratic manner'' spake the LO.
IMHO, `bureaucratic' implies and is implied by `lazy and unprofessional.'
Kindly comment. Yhs.

ahish said...

These resemble astrological predictions in their verisimilitude!

AmOK said...

Is it "forcasting", "forkasting", "foarcasting" or "forecasting"? My spelling is going to the spell-checking dogs, OLO, help me. (Sorry I am a bit late with this one :)).