The reason for this is pretty simple, if the profitability of the network drops 50% in one moment, 50% of the miners should drop off.
Some predict that after halving and no price increase some miners could turn their machines off and hashrate could drop because of that more or less significantly.
Difficulty History
Date Difficulty Change Hash Rate
Apr 01 2016 166,851,513,283 0.82% 1,194,369,655 GH/s
Mar 18 2016 165,496,835,118 4.46% 1,184,672,491 GH/s
Mar 04 2016 158,427,203,767 -3.10% 1,134,066,098 GH/s
Feb 19 2016 163,491,654,909 13.44% 1,170,318,852 GH/s
Feb 07 2016 144,116,447,847 20.06% 1,031,625,717 GH/s
Jan 26 2016 120,033,340,651 5.89% 859,232,121 GH/s
Jan 13 2016 113,354,299,801 9.12% 811,421,684 GH/s
Dec 31 2015 103,880,340,815 11.16% 743,604,444 GH/s
Dec 18 2015 93,448,670,796 18.14% 668,931,642 GH/s
Dec 06 2015 79,102,380,900 8.77% 566,236,898 GH/s
Nov 24 2015 72,722,780,643 10.44% 520,569,941 GH/s
Nov 11 2015 65,848,255,180 5.77% 471,360,171 GH/s
Oct 29 2015 62,253,982,450 2.25% 445,631,364 GH/s
800mn GH/s (out of 1.2bn) were added in the last 5 months. So 2/3rds of the hashpower should, in theory, be representative of state of the art equipment which has an intended lifespan beyond the halving. And there are quite a few months ahead of us for upgrading or introducing more efficient hardware. I think all this equipment is here to stay (and mine). A small hashrate drop is possible though. Larger drops are possible if combined with a price collapse.