Do you have any short fly times? I wrote an application that among other things, can best-fit an exponential curve to speed reduction data. If you send me another data point such as his current 30 meter fly time, combined with his 800 time, it will give a reasonable predictor of where he is currently at in the 400. With only two data points, I use k=0.013, but if you have a couple short distances (e.g. 30m and 60m fly times), it will find his k value too.
Best,
Christopher