With data as noisy as they are and no convincing explicit physical model in the background, eyballing the decrease is (imho) as good as any mathematical model can be. Or in other words: extrapolation of any kind holds only under the assumption, that the processes, which lead to this unexpectedly rapid loss, continue more or less the way they did.
This is not unreasonable.
But on the other hand, a switch to a different weather pattern, leading to, say, more hours of overcast and subsequently less melting, cannot be ruled out.
At the end of the day, we are left with a mixture of common sense, guts feeling and fit functions.