For week 15, we’re looking at a very topical dataset. The National Snow & Ice Data Center has taken daily sea ice coverage measurements dating back to the late 1970’s. Recently, they created the visual below from the data:

What works well

• The time to insight is very low with this visual. I know exactly where the story is
• They cited their source
• The scale is labeled on the y-axis

What could be improved

• I don’t think if the entire legend is necessary because it’s very confusing to see all those years but one only 2 highlighted
• Besides the obvious fact that ice melts in the summer months and freezes in the winter months, I’m curious if there are other anomalies.

So when I first starting playing this dataset, I immediately attempted to recreate this dataset. But then I got curious about each month wondered if there were insights in looking at a median based on month as opposed to year. So that was the driving force behind my final product. Here it is, click for the interactive version.

Each line represents the difference from the median of each month. There are two things that really stick out in view:

• Each month crossed the median line in roughly the same year, 1999.
• But as you can see here, the summer months (July-Sept) are really tipping the scales. And that carries over into the winter months.

So how did I do this?

The first thing I did was calculated the 30-year median for each month. That calculation looks like this:

{ FIXED MONTH([Date]):median(if YEAR([Date])>=1981
AND YEAR([Date])<=2011
THEN
[Extent (million sq km)]
END)
}

After that, I calculated the difference like this

median([Extent (million sq km)])-min([30 year median])

This is the crosstab view one month so we can understand what these calculations are doing

You can see that the 30-year median changes at each month and then the value for each month/year is then subtracted to get the difference.

I hope this helps. Feel free to download my workbook as well and reverse engineer it.

Until next time!