Some Sources of Error in Using Pellet-Group Counts for Censusing Deer

1. What they did

In Michigan, deer numbers were estimated by counting deer pellet groups (piles of droppings) on small sample plots. Deer drop about 13 pellet groups a day, so if every group laid down since autumn leaf fall is found, the count can be turned into the average number of deer over winter. That rests on the assumption that observers find every group and tell new groups from old ones correctly. The authors tested this in a fenced, mile-square deer enclosure at the Cusino Wildlife Experiment Station.

• Counting accuracy. The plots were circles of one fiftieth of an acre, 306 in all, laid out in pairs along a grid. In autumn 1955, before leaf fall and the first snow, the senior author searched each plot several times and mapped every pellet group already there. In spring 1956, two experienced game biologists who had not taken part in the autumn survey counted the groups on 293 plots (88 and 205 plots each; 13 were under deep water) and classed each as old (before leaf fall) or new (after leaf fall). The senior author and an aide then searched each plot again at once and compared the results with the autumn records.
• Persistence of droppings. From summer 1953, they marked 318 pellet groups with numbered stakes (123 laid down in summer, before leaf fall, and 195 after leaf fall) and revisited them each spring or early summer for up to 5 years. Each time they recorded the shape, colour, size and number of pellets, how far they had decayed, and how old they looked, along with the ground moisture, ground cover, shade and overstory at the site.

2. What they found

Counting accuracy

• The biologists counted 682 pellet groups; the recheck showed 869 had actually been dropped since leaf fall. Taken together, the errors reduced the count by 22% of the new groups actually present.
• 43% of the mistakes were groups missed, most of which the senior author considered clearly visible.
• 45% were new groups called old. Many of these were late-autumn droppings made largely of the coarse remains of black cherries, which broke down fast and did not look like normal pellets.
• Smaller errors: old groups called new (6.5% of mistakes), porcupine pellets called deer pellets (2.0%), and one group counted as two (3.3%), mostly on trails or in deer yards where many groups lie close together.
• Errors that lowered the count made up 88% of all errors, so the survey underestimated the deer.
• The two biologists differed significantly in accuracy. For one of them, the share of groups found depended on how many groups were on the plot; for the other it did not.

How long droppings last

• Of the winter (after leaf fall) groups, 53% were still visible the second spring, and the senior author judged that 7% would have been taken for new groups by qualified observers. By the third spring 26% were visible, and none looked new. A few groups under heavy cover were still solid after 5 years.
• Second-spring survival was lowest in swamps (43%) and deciduous hardwoods (50%), higher in mixed hardwoods and conifers (56%) and highest in the open (75%). Groups on wet ground disappeared faster than those on dry ground (27% versus 66% still visible). On dry sites pellets often stayed hard and unchanged for years, and groups in the light lasted better than those in shade.
• Ground cover mattered: at least 58% of groups on grasses, bracken fern, conifer needles, or mosses and herbs were visible the second spring, against less than 52% on leaves or leaves and herbs, probably because of how much dead vegetation is added to the litter each year. Groups on Sphagnum moss vanished by midsummer as the moss grew over them, while those on slow-growing mosses on dry, open sites often lasted for years.
• Of 123 summer groups, 32% were still visible the next spring, and 13% of those looked fresh enough to be taken for new groups. Few lasted to the second spring, and all of those looked old.

3. What they concluded

• The evidence could suggest the pellet-group survey fails before it begins, but the authors do not believe that is altogether true.
• The two biggest errors, missed groups and new groups called old, could be reduced by using a plot shape that is easier to search, by having two experienced observers check each other’s work, and by aging groups by their position on the leaf litter rather than by how they look.
• With intensive instruction that stresses the errors that can be made, the survey can be a useful tool. Michigan now uses long, narrow rectangular plots (72.6 by 12 feet) searched by two people who start from opposite ends and then double-check each other’s halves.
• Even with better methods, the estimate of average deer use would probably be on the low side.

4. How tracking was or was not used

The whole study is about reading one kind of sign: deer droppings. Observers had to find pellet groups, tell deer pellets from porcupine pellets, separate neighbouring groups, and judge whether each group was dropped before or after leaf fall. The persistence study recorded how droppings change and decay over several years in different habitats, which is the basis for aging them.

5. Tracker or observer certification

The two observers are described as experienced game biologists. Their skill was tested in this study, against the senior author’s autumn mapping and spring recheck, and they differed significantly in accuracy. The senior author’s own judgments of what was visible and how old groups looked served as the reference, and were not themselves tested.

6. Other technologies, skills or innovations

• Each pellet group was mapped by compass angle and distance from the plot centre before leaf fall, so it could be checked in spring.
• Groups found in spring were marked with different coloured metal disks for old and new.
• Marked groups were followed with numbered stakes for up to 5 years.

7. Implications for conservation (from the paper)

• Deer censuses based on pellet-group counts are likely to underestimate deer numbers, because missed and misaged groups far outnumber errors that inflate the count.
• Observers can differ greatly, so no valid index values could be calculated from this survey.
• Training that stresses the common errors, double-observer searches and easier plot shapes can make the method more useful.