Sampling strategies

The three sampling strategies used in the virtual forest tutorial were Systematic, Random, and Haphazard, using the area-based method. Timewise, the Systematic method was the most efficient by 1 hour. The other two took a similar amount of time, the Random being slightly faster.

• Systematic 11 h 36 m
• Random 12 h 39 m
• Haphazard 12 h 47 m 

The most accurate system was the Random sampling, followed by the Haphazard method and then the Systematic. Using the formula in “Tutorial: Sampling Theory Using Virtual Forests Tutorial,” I calculated the percent errors for all species based on actual density (Table 1), and by averaging the results, we get the following average percentage errors:

• Systematic 74%
• Random 71%
• Haphazard 72%

All sampling methods were accurate with the most abundant species, but accuracy significantly decreased with lower densities. No strategy was accurate for rare species – all failed to account for three of the rarest species, and the numbers were at least double for the following three. The least bad was the Random strategy.

The Random and Haphazard methods give the most accurate results; however, they take the longest to complete. The average percent errors are high for all sampling methods due to their low sensitivity to rare species. This low precision for rare species leads me to believe that 24 samples are insufficient if the goal is to study uncommon species.

I believe the Systematic strategy did poorly because the results greatly vary depending on the starting point, particularly if there are subareas with different physical or environmental conditions. Although the systematic method has the lowest accuracy, it has the advantage of being the fastest. This could be improved by having various starting points or raising the number of sample points.

Although the Haphazard method did well overall, the results were, well, haphazardous. It did very well on some species but varied wildly on others, giving, on average, a lower percentage error but inconsistent results. For example, it gave the worst estimate for the most common species but, through luck, got the same number for the second most common one.

Overall, if accuracy is critical, using a Random method is the best strategy; however, if time is a constraint, the best approach is to use Systematic sampling.

Table 1. Actual and surveyed density with corresponding percentage error for all species using the three different sampling methods.