Select the month and year (2013 through 2015) to compare tornado occurrence with TEI and SCP indexes; The indexes use CFSv2 realtime products.

For the X month forecast I am using all ensemble members from 3 precedent days. As of now I use a fixed L of 31 days and then average. For February - for example - it means that I am using the average over the first of March. I can be more precise in the selection of the data but I don't think it would make a big difference