They probably have voluntary reporting by commercial beekeepers. Figured they had to do something to monitor the situation with all the hysteria of "AIIEEE! The bees are all dying and we're all going to starve to death!!"
Regardless of just how they determine the level of hives lost, it's obvious that the 27% decline in hive loss from the previous year can't be attributed to a reduction in usage of glyphosate, since usage of that herbicide actually keeps going up each year.
That makes sense. However, if I posted a statistics from a 1 year trend, you would rightly object that one year's statistics could be statistically dubious. Also, there could have occurred phenomena that reduced the numbers of bees' natural enemies or microbes that infect bees that year.
If the trend continues, that should cast doubt on the glyphosphate/bee death hypothesis.