For the virtual forest sampling tutorial I used the area method. Systematic sampling was the fast approach (12 hours, 36 mins) followed by Haphazard sampling (13 hours, 5 mins) then Random sampling (13 hours, 12 mins). I had a feeling random sampling would be the slowest just because the surveyor would have to go randomly all over. I actually ran two instances of haphazard sampling, one where I “sampled” from west to east (from top to bottom) of each section (basically spread out my sampling with a zigzag pattern from the top to bottom of the screen in no particular placement) and the I did it again which more “lazy” approach were I focused on the east portion of the area and basically targeted the more dense patches. The first haphazard approach was definitely longer, but the second “lazy” approach was only 4 minutes longer than the Systematic sampling method. So it’s evident that the Haphazard approach would differ in sampling time depending on the surveyors approach. For the purpose of this discussion, I used the results from the first Haphazard method (zigzag pattern from west to east in each of the five topographical categories).
The 2 most common species found in my sampling were Eastern Hemlock and Sweet Birch, and the 2 least common species were Striped Maple and White Pine. The accuracy between the most common and the rarest species did change but not consistently between the 3 methods. For the Haphazard sampling, the 2 most common species had lower % errors than the 2 rarest species – this is what I expected for results across the board. For the other 2 methods (systematic and random) the average % error was overall less for common species and higher for the rare species; however, Striped Maple (rare species) had an individual lower % error than Sweet Birch (common species).
Interestingly, the Haphazard method had the overall lower average % error for all species except Striped Maple. I wonder if my Haphazard sampling approach covered more of the plot, and because we had 5 sample locations in each of the five topographical regions, it resulted in a better representation of the plot as a whole. Whereas the systematic sampling only covered a limited strip of land along the east side of the plot, and the random sampling perhaps didn’t provide an even coverage of the plot as a whole or an equal representation of the five topographic regions.
| Species | Actual Density | Area; Systematic sampling density | % Error | Area; Random sampling density | % Error | Area; Haphazard sampling density | % Error |
| Eastern Hemlock | 469.9 | 412 | 12.3 | 388 | 17.4 | 476 | 1.3 |
| Sweet Birch | 117.5 | 72 | 38.7 | 72 | 38.7 | 124 | 5.5 |
| Yellow Birch | 108.9 | 136 | 24.9 | 100 | 8.2 | 100 | 8.2 |
| Chestnut Oak | 87.5 | 80 | 8.6 | 96 | 9.7 | 84 | 4.0 |
| Red Maple | 118.9 | 104 | 12.5 | 96 | 19.3 | 108 | 9.2 |
| Striped Maple | 17.5 | 16 | 8.6 | 12 | 31.4 | 20 | 14.3 |
| White Pine | 8.4 | 0 | 100.0 | 24 | 185.7 | 4 | 52.4 |
I quite like this exercise. I think it shows the value of the different sampling approaches.
Haphazard can indeed cover more of the plot. Haphazard by nature will cover more of the plot but will be less predictable in what it is gathering.
Systematic sampling – If you are using it to sample one type of something (say a certain plant community), it can be fairly/relatively clustered, as long as you are certain that it is representative of that type.
if systematic sampling was sampling in a habitat type that favours striped maple (like an edge, which it did) then you might get what you see.