TRH015, installed at 2m on Aug 12 18:51 UTC also reports
incorrect humidities. Its cutoff is around 45%. Above 45%
the reported humidity is anti-correlated with the 4 other
sensors - looking more like a temperature.
In this plot, the night of Aug 11-12 shows the problem with
unit 013, and the next night shows the problem with 015:
http://www.eol.ucar.edu/isf/projects/BEACHON_SRM/isfs/qcdata/plots/20090812/trh_20090812.png