The plugging of highway culverts by beavers (Castor canadensis) creates roa
dside impoundments that damage and sometimes flood the roadbed. Continually
mitigating these problem sites requires considerable time, money, and reso
urces from town, county, and state highway departments. We initiated this s
tudy to develop proactive and long-term approaches to deal with nuisance be
avers along roadsides. Our specific objective was to compare culvert and ha
bitat features at at plugged and nonplugged culverts. From June to October
1997 and 1998, we sampled 216 roadside sites in New York state: 113 sites w
here beavers plugged the highway culvert and 103 sites where beavers did no
t plug the culvert but instead constructed an upstream or downstream dam. W
e used stepwise logistic regression (SLR) to identify key variables associa
ted with plugged culverts. We evaluate classification rates of regression m
odel with measures of sensitivity and specificity. For the combined data se
t, the logistic function retained culvert inlet opening area (m(2)) and str
eam gradient in the final model. Based on these 2 variables, the model corr
ectly classified 79% of the sites. Our results indicated that installing ov
ersized culverts would have the greatest influence on discouraging beaver p
lugging activity. Prorated over the service life of culverts, the installat
ion of oversized culverts by highway departments may be more cost-effective
than trapping, debris removal, or other short-term options to manage beave
r damage to roads.