SQL - it is very handy!

SQL is really very useful. This single query allows me to find the greatest range of a particular variable for every day in my database (which is over 5000 days).

Now, looking at the results, I'm a little sceptical of the result for 10th February 1998, surely the -99.9° is an error, but, regardless of this fact, it's still quite informative. Currently I'm interpolating the air temperature across France, and I've run into a bit of a quantization problem. The average air temperature over my data range is about 13° but clearly, the temperature that I estimate at each pixel is a floating point value, which when set as the intensity of the pixel, gets rounded to an integer. So, on days where the range of temperatures is fairly low, the resultant intepolated image is completely homogeneous, because the values all round to the same number.

Clearly we need to perform some sort of histogram equalisation, to minimise the quantisation error, yet also somehow ensure that the actual temperature can still be read from the image... Umm...