Statistical Evaluation of Remotely Sensed Thermal Data for Deer Census

Abstract

A statistical decision theory that originated with the development of radar and electronic communications systems was adapted to the problem of remote detection of white-tailed deer (Odocoileus virginianus) and mule deer (0. hemionus). The theory provided a model by which thermal emission data obtained in a ground-based field study of the effective radiant temperature of deer, brush, and snow were used to evaluate feasibility of use of airborne thermal scanners for deer census. Deer could be detected successfully against a snow-covered background under conditions existing during 2 widely separated field experiments. However, a thermal scanner would exhibit large errors in detecting deer when the probability of encountering snow-free objects was greater than the probability of encountering deer. Thermal contrast, by itself, is not of value in censusing deer populations. J. WILDL. MANAGE. 44(2):397-402 Thermal infrared (ir) scanning techniques of remote sensing have been proposed for censusing white-tailed and mule deer. Most studies were conducted in the spectral ranges of 3-5 or 8-14 gm. Choice of the thermal emission bands appeared to have been based upon the assumption that systems developed for military use might be applied to wildlife detection, because an animal, like a military target (tank, jet plane, etc.), was a heat-generating system. Croon et al. (1968) and McCullough et al. (1969) speculated that white-tailed deer might be censused with an aerial thermal scanner in the 8to 14-Aum range. Graves et al. (1972) found the 3to 5-am range was best under certain conditions. Parker and Driscoll (1972) and Parker (1972) concluded that thermal scanning for wildlife detection was feasible, but research was needed to optimize a thermal scanner for deer detection. Isakson et al. (1975) used digital techniques to interpret thermal ir data; results were superior to those obtained by visual interpretation of ir imagery, but were experimental and not operational. Wride and Baker (1977) at Elk Island Park, Alberta, Canada, reported limited success with an operational system, but absence of ground truth made the value of the experiment difficult to