Signing of non-permanent road surface conditions, such as ice, is difficult
because hazard formation, location, and duration are unpredictable. Subseq
uently, many state transportation departments have begun to question the se
nsibility of expending material and personnel resources to maintain ice war
ning signs when little proof exists of their effectiveness in improving hig
hway safety. This research statistically studies the effectiveness of ice w
arning signs in reducing accident frequency and accident severity in Washin
gton State. Our findings show that the presence of ice warning signs was no
t a significant factor in reducing ice-accident frequency or ice-accident s
everity. However, we were able to identify significant spatial, temporal, t
raffic, roadway and accident characteristics that influenced ice-accident f
requency and severity. The identification of these characteristics will all
ow for better placement of ice warning signs and improvements in roadway an
d roadside design that can reduce the frequency and severity of ice-related
accidents. (C) 2000 Elsevier Science Ltd. All rights reserved.