Detecting river otter (Lontra canadensis) presence or estimating abundance
relies on harvest records, trapper surveys, track surveys, or latrine surve
ys. Harvest records and trapper surveys are not an option where otters are
protected, and track surveys have limited utility in many areas. Latrine su
rveys are often useful, but may be labor intensive. We used multivariate an
alysis techniques to examine habitat characteristics at 131 river otter lat
rines and 113 randomly chosen (nonlatrine) sites along upper Pine Creek, no
rthcentral Pennsylvania, 1991-1992. Discriminant analysis and logistic regr
ession each identified 6 variables as predictors of otter latrine sites: ve
rtical banks, rock formations, points of land, backwater sloughs, tributary
streams, and beaver (Castor canadensis) bank dens, lodges, or ponds. Model
s were cross-validated with ground surveys and low-altitude aerial photogra
phs (1:1,300) from lower Pine Creek, 1993-1994, and Tionesta Creek, northwe
stern Pennsylvania, 1993-1994. We developed a pattern recognition (PATREC)
model using the 6 variables identified as predictors of river otter latrine
s. Results were similar for all 3 model types, but differed among the 3 sur
vey areas. All 3 model types showed potential for identifying latrines. All
ocation of resources to detect otter presence can be adjusted by varying th
e cutpoint P(E), that defines a predicted latrine for both the logistic reg
ression and the PATREC models.