Bill -

That comes out at a loss rate of roughly -600k sq kms/decade. (For comparison, the equivalent NSIDC value for the period 1979 until now is about -542k sq kms/decade.)

As the last period (drop of 633 sq km in just under 7 years) - I made it loss rate of around 638k sq kms/decade - and accelerating. Or have I missed something?

Charles, for that "

*roughly -600k sq kms/decade*" figure from the IJIS/JAXA/ADS dataset, I didn't even bother doing any formal calculation. The 1980's average (11.957 million sq kms) was just over 1.8 million sq kms above the 2015 year-end average of 10.111 million sq kms. Those two "end-point" numbers imply that each year - when averaged across the dataset and assuming a linear trend - the annual average drops by just over 60k sq kms.

The point I was trying to make, but clearly failed to adequately do so, was that the current level of annual average extent (~10,000k) is more than of an order of magnitude in excess of the decadal drop rate (600k). For such a purpose, the difference between 600k and 638k is largely irrelevant. (Although I totally agree that this value is increasing - and will continue to do so.)

I should have been more explicit in placing this in the context of a suggestion that has been made on this thread, namely that the annual average could drop to Zero within 6 or 7 years. For this to happen, the

**annual** drop,

**averaged over the next 6 or 7 years**, would need to become almost as large as the

**total** drop seen thus far.

The point of the first chart (comment #3524) was to show that the trend swings wildly up and down, and that therefore it is something of a folly to say that "this time it's for real, and we can project this particular value of the trend into the future". The huge drop seen in the 1-year rolling annual average in 2007, and the gentler but deeper drops in 2011 and across the 2012/13 periods failed to show any staying power - so why should this time be different?

The purpose of the second chart was to show just how much there is still to go before we get anywhere near a zero annual average.

Had I been aiming to produce a trend value with greater precision (but not necessarily of greater accuracy) I would used my old faithful friend - Excel's SLOPE function. However, when one is working with a mixture of both annual and decadal averages, it becomes necessary to give some consideration to just where along the X-axis the various data points appear. Normally I treat annually averaged data as occurring on the middle of the relevant year. However, for consistency, the decadal average should then be treated as occurring at the boundary of the 5th and 6th year. For example, the 1980's average would be treated as being located at 31st Dec 1984 (or 1st Jan 1985). As the 2015 data point should be tied to the mid-year, it means that there is effectively a 30.5 year range spanning the 1980's average and the 2015 average.

As the data within the .csv file only has decadal values for its early years, I felt that using SLOPE would perforce attach a level of "legitimacy" to any result that the granularity of the early years simply did not justify.

I added the NSIDC figures merely to show that these were pretty similar to the "rough & ready" value I had just obtained using the JAXA/IJIS data.

For the NSIDC figures, I used their monthly dataset from...

ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/To obtain an end-of-year annual average, I simply averaged the 12 relevant monthly values. (

*I know that's only an approximation, and that February in particular is therefore excessively weighted. However, as this effect applies to each year, it tends to cancel out when one is looking for a trend in annual averages values.*) This simple approach was adopted purely for expediency in this case. On another spreadsheet that I use, I have weighted each month according to its actual number of days. However, the difference is pretty marginal.

I then used Excel's SLOPE function on the 1979 - 2016 annual averages thus obtained to derive the drop rate of around 542k sq kms/decade. (

*NB As 2016 is obviously not over yet, I plugged in a stop-gap value for the most recent 12 months, i.e. Dec 2015 - Nov 2016. As December this year is looking as though it will average out considerably less than Dec 2015, that decadal drop rate figure will probably end up being closer to 548k sq kms.*)