Dietary Analysis from Fecal Samples: How Many Scats Are Enough?

1. What they did

Many studies work out what mammals eat by identifying the hard parts of prey, such as bones, fish ear bones, insect shells and seeds, that pass through the gut and end up in scats. Researchers then count how often each prey species turns up across a set of scats (its frequency of occurrence) and compare diets between places, seasons, years, sexes or species. Published studies have used anything from 10 to more than 1,000 scats, and the authors asked how many are really needed.

• Computer-made scats. They generated pairs of scat populations containing 3 to 15 main prey species, each occurring in at least 5% of scats. One population of each pair had the same frequency for every species; the other had frequencies that declined linearly, declined exponentially, or were equal except for 1 dominant species. Each pair differed by a moderate amount (an effect size of 0.30).
• Repeated sampling. For each scenario they drew 2 random samples of 10 scats, tested whether the diets differed, and repeated this 50,000 times. They then raised the sample size in steps of 10 up to 200 scats. The minimum sample size was the smallest number of scats that detected the difference at least 80% of the time.
• Real scats. They repeated the exercise with Steller sea lion scats collected in Southeast Alaska in the summers of 1997 and 1998, comparing scats from breeding areas used by females with scats from a haulout used by males. They also pooled the prey into 8 categories and ran it again.

2. What they found

Simulations

• Fewer prey species in the diet meant more scats were needed. The authors note this may seem counterintuitive: at the same effect size, the gap between 2 diets with few species is smaller than between diets with many species.
• If prey frequencies decline exponentially (the pattern most often seen in real scat data), 51 scats were needed with 15 prey species, 94 with 6 species and 179 with 3 species.
• If frequencies decline linearly, 23 scats were needed with 15 species and 168 with 3 species.
• The most extreme scenario (equal frequencies except 1 species) needed about 106 scats with 15 species and more than 200 with 3 or 4 species. The authors consider this pattern unlikely in the wild.
• Based on binomial probability, at least 59 scats give 95% confidence of finding at least 1 scat containing a prey species that occurs in 5% of scats.

Steller sea lion scats

• Females’ scats held 22 prey species and males’ scats 26 (31 in all). Only 7 species in females’ and 10 in males’ scats occurred in at least 5% of scats; 11 species overall.
• The real difference between male and female diets was larger than moderate (effect size 0.42), so only 19 scats were enough to show that their diets differed. Pooling prey into 8 categories gave a similar answer (effect size 0.39, 18 scats).
• Had the difference been moderate (0.30), about 85 scats would have been needed.
• Sample size did not change the rate of false differences when male scats were compared with each other.

3. What they concluded

• About 60 scats is a reasonable rule of thumb for finding which prey are present in a diet, and is enough to detect large differences.
• About 100 scats (from 1 site, or pooled across sites) is more appropriate when the aim is to detect moderate differences in diet over time or between areas; 94 was their conservative minimum for diets of 6 or more principal prey species.
• These numbers are larger than many would expect because the authors chose the most cautious scenarios, so that real differences in diet are not missed.
• How scats are collected matters as much as how many: a large set from a few individuals, 1 sex, 1 site or a few days may be badly biased, and a thoughtfully collected smaller sample may represent a population’s diet better than a larger haphazard one.
• The simulations ignored digestion rates, scat size and prey size. Captive feeding trials are needed to understand what the remains in scats actually represent, and could change the recommended numbers.
• The results also apply to other diet data, such as stomach contents.

4. How tracking was or was not used

No tracks were followed or identified. The study is about scats as a source of diet information: the field scats were collected at sea lion breeding areas and haulouts in Alaska, and prey were identified from hard parts in the scats. The paper notes that some species bury their scats while others leave them in predictable places, and that sea lion scats may build up above the high-water line while harbor seal scats are often washed away by the tide each day.

5. Tracker or observer certification

Not reported. The paper does not say who collected the scats or identified the prey remains.

6. Other technologies, skills or innovations

• Monte Carlo simulation of scat samples to estimate the sample size needed before a study starts, when the real diet is not yet known.
• Identifying prey from all hard structures in a scat (not only fish ear bones) reduces the problem of bones being fully digested.