The location of wildlife is frequently determined using telemetry data
gathered at short intervals. If radio transmissions are reflected, as
often occurs in mountainous regions, then existing location estimatio
n techniques are unreliable. We explore the effects of gross observati
on errors upon current analyses and suggest an alternative analysis ba
sed on robust state-space time-series modeling. We determine location
estimates and their precisions, both for simulated and real mule-deer
data, using current and robust procedures. Implementation and specific
ation of filter parameters are also discussed. We conclude that the pr
oposed filter-smoother is similar to the Gaussian filter-smoother when
data are not greatly contaminated and that the robust version improve
s upon location estimates when contamination is large.