User: | Open Learning Faculty Member:
My initial sampling day went as planned insofar as I was able to collect data using the method of walking a transect and placing my quadrat after a random number of paces. I was even able to find my target species in some of the quadrats and record useful data. My sampling plan was flawed, however, as it assumed a higher density of dog strangling vine. I ended up covering the whole length of my study area before I was able to obtain 5 quadrats containing the target species. The issue is that, while abundant, the vine grows in a small number of patches. My initial design attempted to avoid sampling more than once from each patch by having a minimum number of paces, but this turned out to be a problem as there are only 4 patches in the particular treatment area I was sampling. Further investigation revealed the same issue in the other treatment areas. My study area doesn’t have enough Dog Strangling vine patches to be sampled in this way.
The response variable data collected was generally as expected. The number of seed pods on the plants reflected the different treatment areas as predicted in my hypothesis. The soil moisture or predictor variable measurements were not as expected. All of the soil was measured to be dry, regardless of the treatment area. During the sampling it became apparent that this was another flaw in the study design. To demonstrate that the slope or manicured areas had less water than the area along the creek (which seems to be the case based on the compaction and visual dryness of the soil), I would need data throughout the growing season that showed the plants consistently had less water. A simple snapshot would not demonstrate this effectively.
I am going to modify my sampling approach in two ways.
First, since my sample unit is the individual plant, I am going to sample the various patches in greater depth instead of using random paces and quadrats along a transect. This will provide several replicate areas within each treatment and allow for a larger number of individuals to be counted. A quadrat will be used within each patch to collect 3 or 4 separate samples, several meters apart from each other. A random number generator will be used to determine random locations within each patch for sampling.
Second, since I can’t demonstrate that water availability is different across time between the different treatment areas. I will have to rethink my hypothesis. The data supports the observation that there is a gradient in the number of seeds between the different treatment areas so the predictor variable will be modified to be the different treatment areas themselves within the study area.
These two modifications will allow for significantly more individuals to be sampled within the study area while maintaining random selection, and allow for a hypothesis that is verifiable or refutable.
