The cadastral mi2 (2.59 km2) was evaluated as a primary sample unit (SU) of a stratified 2-stage sampling system used to estimate densities of mule deer (Odocoileus hemionus) pellet groups occurring on juniper-pinyon winter range in the Piceance Basin of Colorado during 1976 and 1977. In preliminary sampling in 1976, 26 SU systematically subsampled with a grid of 100 0.001-ha temporary circular plots generated a 90% confidence interval (CI) of ?18% about new groups per SU. From optimal allocation of 1976 data, we projected 23 SU or 4% of the potential SU subsampled with 100 plots would produce a 90% CI of ?10%. Alternative subsampling plans consisting of subsets of the 100-plot grid indicated subsampling could be reduced. In 1977, an alternative subsampling plan was used and the 90% CI was ?14% about new groups per SU. The 2.59-km2 SU combined with subsampling effectively estimated densities of pellet groups and should be considered when designing sampling systems. J. WILDL. MANAGE. 47(2):476-485 A common approach to estimating sizes of ungulate populations involves converting fecal pellet-group densities into numbers of animals (Bennett et al. 1940, Riney 1957, Robinette et al. 1958, Smith et al. 1969, Ryel 1971). Reliability of the technique has varied, but limited evidence indicates that pellet-group counts can produce reasonable estimates of animal numbers (Eberhardt and Van Etten 1956, Ryel 1971). Here we report on a stratified 2-stage sampling system used to estimate densities of mule deer pellet groups. A major problem encountered in estimating densities of pellet groups is designing sampling and estimation procedures that have acceptable levels of precision. This is especially evident when sampling large geographic areas such as winter ranges of cervid populations. The sample unit is fundamentally important to any sampling system. Sample units commonly used in pellet-group surveys include individual plots or belt transects (Wallmo et al. 1962, Neff 1968). Alternatively, Ryel (1971) used cadastral mi2 (2.59 km2) sample units and subsampled these primary units with 1 transect containing 8 0.008-ha ('/5o-acre) plots. Batcheler (1975) discarded bounded plot sample units in favor of sampling distances between pellet groups. Frequency distributions of pellet groups enumerated on bounded plots have been described by the non-normal negative binomial distribution, indicating a contagious distribution of pellet groups (Bowden et al. 1969, McConnell and Smith 1970, Ryel 1971, Stormer et al. 1977). As a result, sample variances based on plots tend to be large, necessitating large numbers of plots to generate acceptable precision. Several factors influence variation in numbers of pellet groups found on plots. Terrai and vegetation affect distributions of pellet groups and likely influence the common occurrence of many sample plots containing no groups and few plots having m ny groups (Harris 1959, Anderson et al. 1972). Also, probability of errors in counting pellet groups per plot increases with increasing density of vegetation or numbers of groups on plots (Smith 1968). This study was undertaken to further evaluate the cadastral mi2 (2.59 km2) as a 476 J. Wildl. Manage. 47(2):1983 This content downloaded from 207.46.13.120 on Wed, 14 Sep 2016 04:15:02 UTC All use subject to http://about.jstor.org/terms SAMPLING PELLET-GROUP DENSITIES * Freddy and Bowden 477 primary sample unit (SU) for sampling pellet-group densities. In 1976, a preliminary sampling of mule deer pellet-group densities in the Piceance Basin of northwestern Colorado was conducted. From this preliminary sample, a modified sampling system was designed and tested in 1977. We specifically address questions of sampling intensity and precision, alternate subsampling strategies, logistics of sampling, and estimates of deer population size based on our samples. We thank A. E. Anderson for assistance in developing methodology and seasonal employees for collecting data. R. B. Gill, N. T. Hobbs, and D. F. Reed constructively reviewed the manuscript. This study was supported by Colorado Federal Aid in Wildlife Restoration Projects W-38-R and W-126-R.