There have been intense rains in Chennai for the past few weeks and when not mopping up the pieces, everyone is wondering whether we are getting more rains every year. I tried looking for historical rain data/ graphs in TN and Chennai to see if this year was truly an anomaly. With a little bit of digging, some data was available.

This site provides data on rainfall in mms in Tamil Nadu and Pondicherry from 1871 to 2013, month-wise. These are some quick plots from that data-set. First, the total rainfall per year and the running average of that for 10 and 25 years. It does seem like there is an up-tick in mm of rain in both the smoothed lines, especially in the 10 year running average.

The histogram of the rainfall data is close enough to normal (eyeballing) to think about standard deviations from the mean.

So, we ask the question, “When are the years with ‘abnormal’ rainfall?”(<5% probability, outside shaded region, more than1.96 SD). The first 100 years (1871-1971) look like you would expect: 6 such events, 3 at the high end and 3 at the low end. However, in the 44 years since, there have been 5 events, all with ‘too much’ rain. 3 of them in the last 10 years since 2005.

From this (admittedly quick and hand-wavy) look at the data, I’d bet that Chennai (TN) is getting a lot rainier than in the past. We are in the beginning of a period with increasing rains, so our expectations for what is a ‘normal’ amount of rain should change. It is also possible that the distribution of rainfall is wider, which could mean more years with failed rains also. Either way, this indicates that the last couple of decades and the coming decades are qualitatively different from the century before that.

Some caveats on data: The Indian Institute of Tropical Meteorology divides the country into subdivisions and collates weather data (read me). TN and Pondy fall into a subdivision and as you can see, the subdivisions fall cleanly along state lines. Whereas, northern TN rains are more like Rayalaseema than Thirunelveli -so this is a bit of an approximation. (The site has Rayalaseema data as well).

I have summed rainfall over each calendar year. I think this is a less-bad approximation, as in this region, most of the rain falls in October, November and December (though the season may extend to February.) In fact, in the data-set, the month of maximum rainfall over a calendar year has always been November or December except for 6 times in ~140 years. (It is September on those years.)

Also, the data stopped at 2013. So I ‘made up’ data for 2014 and 2015. 2014, I assumed was an average year for the area (rain = average of past 143 years) and in 2015, rain = average + 2.1*SD. The rains have affected large parts of northern and central TN, so I have assumed that this year is a large deviation for the entire state. The conclusion holds true even if I hadn’t added the ‘made-up’ data.