In the virtual forests tutorial, I virtually sampled the Snyder-Middleswarth Natural Area using area-based methods in three approaches: systematic, random, and haphazard sampling. The approach with the fastest estimated sampling time was systematic (12 hours, 37 minutes), followed by random (12 hours, 44 minutes), and haphazard (13 hours, 6 minutes). We can compare the percentage error of the different sampling techniques for the two most common and least common species as follows:
Most common species (eastern hemlock):
Percentage error, systematic: (416 – 469.9) / 469.9 * 100 = 11.4%
Percentage error, random: (466.7 – 469.9) / 469.9 * 100 = 0.7%
Percentage error, haphazard (540.0 – 469.9) / 469.9 * 100 = 14.9%
Second most common species (sweet birch):
Percentage error, systematic: (92.0 – 117.5) / 117.5 * 100 = 21.7%
Percentage error, random: (125.0 – 117.5) / 117.5 * 100 = 6.4%
Percentage error, haphazard (208.0 – 117.5) / 117.5 * 100 = 77.0%
Least common species (white pine):
Percentage error, systematic: (16.0 – 8.4) / 8.4 * 100 = 90.5%
Percentage error, random: (8.3 – 8.4) / 8.4 * 100 = 1.2%
Percentage error, haphazard (4.0 – 8.4) / 8.4 * 100 = 52.4%
Second least common species (striped maple):
Percentage error, systematic: (0.0 – 17.5) / 17.5 * 100 = 100%
Percentage error, random: (25.0 – 17.5) / 17.5 * 100 = 42.9%
Percentage error, haphazard (8.0 – 17.5) / 17.5 * 100 = 54.3%
The most accurate sampling strategy for the two common species was random sampling, with an average percentage error of 3.55%. The most accurate sampling strategy for the two rarest species was also random sampling, with an average percentage error of 22.05%. Clearly, accuracy decreases with species abundance, due to the bias introduced by low sample sizes and the relatively low number of sample points inherent to our sampling (i.e. 24 data points over a fairly large area).