If we keep this slope, being on the trend would be enough in 2018 to break the record of 2012.
Etienne
It can be interesting to watch slope developments. Back in the early months of 2010, I vaguely recall doing something similar with the 1979-2009 NSIDC September averages.
Using the then available subset, the trend line did not hit the 2007 jaw dropper until 2024. As one might expect, statistical outliers (~ 2 Stnd Dev) such as 1996, 2007 and, more recently, 2012, can wreak havoc with seemingly relatively stable trends (or the first derivatives thereof).
When one expands the data set to cover 1979-2015, the trend line (virtually) reaches the 2007 level by 2020. Extrapolating further, that 1979-2015 trend line reaches 2012 proportions by 2028.
However, using a rolling annual average smooths out more of the wild fluctuations. Consequently, as your projection shows, any extrapolation of the trend line would not take as long to reach the record low set for the period straddling 2012 and 2013. (If anyone is uncertain about the 2013 bit, that year had lower extent than 2012 over the first few months.)
++++++
Can I also echo the sentiments of so may others in wishing Wip all the best, and thanks ever so much for all your informed commentary!