To start with, I needed a compare period. For that I used a 30-year period preceding this heat wave, so I took the data from January 1, 1980 - December 31, 2009. To determine the compares, I simply averaged each day's high temperature in the period from May 1st to September 30th.
These data were extremely noisy, so I smoothed it by representing each day using the week centered at that day. Thus...May 1st in the compare is April 28th, April 29th, April 30th, May 1st, May 2nd, May 3rd, and May 4th from 1980 + the same for 1981 + ... and then averaged by the number of years.
Next, I needed the data during the heat wave. I did the same thing except using 2011's data.
Finally then, the visualization is just the 2011 data minus the compare data at each lat/lon value and for each day. I picked three cities that were strongly affected by it as references and added those on the left. It's pretty obvious from their plots that the temperatures stayed above the compare for 3-4 months straight and it was brutal for all involved.
I settled on these four months by balancing including enough time to show the heat wave against keeping the gif size reasonably small.
If you would like to compare with a normal year, I did the same thing except that I used 1995's data:
I put this visualization together in MATLAB and the data are from Berkeley Earth. If you would like to read more about this heat wave, it was epic enough that it has its own wikipedia page. This is probably the favorite data visualization exercise I've gone through so far. I will almost certainly repeat this with other legendary heat waves in the near future.
No comments:
Post a Comment