Category: Post 5: Design Reflections

Post 5- Design Reflections

TRU Open Learning 1 Comment

I performed the trial data collection a few days ago and found a few kinks in my design. I realized there a confounding variable in one zone (the second growth Doug Fir forest). There were a few mountain biking/dog walking trails that fragmented the area and the plot data that I collected in this section of the forest was completely absent of any ungulate activity at all (no old trails or anything), which is abnormal in my experience. As I continued to collect data and struggle over all the fallen trees and through the overgrown area I noticed that after I navigated across the last trail there was an increase in ungulate activity. I decided that when collecting my data for the final research I would randomize the points in an equally large 5 ha area south of this last trail (where there are no intersecting trails) in order to minimize the influence of human activity on the use of the area by deer and elk.

I also encountered some unexpected difficulty in accurately estimating percent cover of vegetation in the circular plotting technique that I decided upon. I tried my best to use the BC Ministry of Forest “Field Manual for Describing Terrestrial Ecosystems” and follow their guidelines for estimating in a circular plot and I will do my best to be consistent, but I imagine that by my last plot of the 30 replicates will be much more accurate than the first.

I found the randomization of points by GPS coordinates to be well suited for this study design, and I found that the AVENZA maps app on my Iphone to be easy to follow and navigate to the GPS point locations (though an external battery bank charger is necessary for extended amounts of time). The only issue was the accuracy, the GPS was only accurate to one decimal place of one second of latitude and longitude which theoretically should result in a 10×8 ft area for me to then have to try to randomize a location where the plot should go, but the GPS was jumping all over the place and I ended up having to estimate. I couldn’t just put it in the closest pile of deer scat or in the middle of a trail, and placing the plot in the middle of a bush wasn’t something that I would consider naturally. So, I decided I had to find the approximate boundaries of the points area and then estimate the middle so that I wouldn’t be biased in choosing a point.

I had intended on creating an inventory of all the shrub species that I found in each plot but quickly realized that there were species present that I couldn’t identify at this stage of the season. I decided to just enter them as other shrubs and not spend an eternity looking for defining physical characteristics.

One unexpected hurdle I did not expect was the compromising of my safety by choosing to navigate alone to GPS points in very thick bush where black bear and cougars have been recently sighted. I am well acquainted with traversing tough terrain from many seasons of tree planting, logging on the coastal mountain sides, and many extended hunting trips into the back country, so I did not really expect to experience some slight fear as I was half stuck in an Alder thicket looking down at fresh black bear prints following fresh fawn prints in the mud. I decided that I would forgo following this bear and possibly dead fawn into the next thicket just to collect my point data, and I would generate another random point instead. I decided to bring my bear bangers and some bells for my bag when I collect all my data so I don’t almost sneak up on a bear and its kill again.

Post 6

TRU Open Learning Leave a comment

Since my last posts, my project has changed substantially.  I am now observing if the presence of red-osier dogwood within 1 meter of conifer tree decreases its chances of being browsed by ungulates.

This was completed at LC Gunn park in Prince George.  I created seven 30 meter transect lines, of which I counted every regenerating conifer stem within 1m on either side of the transect.  I then recorded the height of the tree, if it was browsed by an ungulate, and if red-osier dogwood was present within a 1 meter radius of the tree.

I actually did notice that less regenerating conifers were browsed if red-osier dogwood was present.  However, I am still unsure as to why this is.

Data collection went smoothly, and I did not see any reason to change my methods.  In total, I counted 68 trees, with the majority being subalpine fir, then Douglas-fir, and hybrid spruce, respectively. Of those, 30 had red-osier dogwood present within 1 meter, while 38 did not.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

The only difficulty that I had implementing my stratified random sampling technique for identifying shorebirds within each of my specific locations (3 locations along a gradient of human presence) was the fact that I didn’t have a way of measuring a large area to ensure that each quadrat/plot (my sampling unit) could be equal in area. I tried to identify birds within a relatively uniform circumference for my sampling plots/qaudrats. The data I collected was not surprising as it seemed to follow the expected trend of increasing shorebird diversity with a decreasing amount of human presence/urbanization. I plan to use the same sampling approach but I would like to collect more replicate data at various times of day to account for any differences in shorebird diversity across time rather than just space.

Blog Post 5 – Design Reflections

TRU Open Learning Leave a comment

Blog Post 5 – 12-02-20

Collecting the initial data in Module 3 proved to be a difficult task. I spent much time trying to ensure that I correctly identified the vegetation species in the current winter conditions. The species I selected were large enough for me to clearly identify what grouping they belonged to and once that task was complete I was able to get a closer look at them to determine what species of tree or bush they were. The sampling strategy I chose for that data collection was Stratified Random sampling. The study area of the park was divided by strata into park land, which the area was mostly comprised of, and pond land, which was a much smaller portion. After diving the two strata, I randomly selected four areas in the park land and sampled the vegetation within, finding this process to be rather difficult for the aforementioned reasons. The two pond areas were relatively easy to measure, especially as I drew closer to the water with much of the regular vegetation diminishing in these areas. The data that I collected seemed to be relatively unsurprising with some of my general expectations having been met. For example, I had made the prediction before sampling that vegetation such as the Broadleaf Cattail (Typha latfolia) would be found only in pond location quatrats. Upon sampling, it was made clear that this prediction was correct. Overall there were no unusual data collected that caused me any surprise. I think from completing this first sampling data collection, I will continue to use the same strategy of Stratified Random sampling as it appeared that this strategy allowed me to accurately find both the abundance and distribution of several species in the park.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Shannon Myles

February 3rd, 2020

 

During my data collection in the field, the systematic sampling strategy proved to be efficient at surveying the area. The few difficulties I encountered during the sampling did not damage the quality of my data in any way. First, the determination of transects was simple, but keeping that transect straight as I collected my subsamples across the field seemed to be a challenge. For the last three transects, I established three or four checkpoints along each transects in order to keep me straight. Having closer targets greatly improved the quality of my transects. Secondly, making my way along a transect turned out to be slightly more challenging than I expected. The vegetation got pretty dense in some portions of the field. I always managed to make my way through it but I had to push through some plants and small shrubs. Applying the quadrat down never was an issue. I would simply drop it over the vegetation of the area, however tall or dense that was.

The data was not surprising to me. These first samples even seem to play in favour of my initial hypothesis – more flowers appeared as I sampled away from the beach. One noticeable aspect of my data was that all types of flowers seemed to be displayed in clusters.

I think that my systematic approach to survey the site was the best option. The data collection was performed with minimal difficulties that were all overcame to maintain the essence of the systematic method. It eliminates the possibility of bias, and more samples will only add to the reliability of my data.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Did I have any difficulties implementing my sampling strategy? Yes!

  • I manufactured a simple 1mx1m frame from cardboard in order to accurately find my quadrats. I will say that a cardboard frame is not the best piece of equipment when its a rainy, gusty day at the beach…
  • The tide was all the way in when I visited the site, which made access to the lowest gradient quite difficult at times.
  • I found access to one of my gradients (the slope with boulders) to be more of a problem than I expected. It was physically difficult to reach the quadrats I was to sample in this area, and I had to resort to looking from a distance which was difficult at times due to the blackberry, snowberry, and other plant material in the same area.

Was the data I found surprising? Yes.

  • Yes, only in that I did actually find a single juvenile Alnus rubra when I expected to find none.

Do I need to modify my approach? Yes.

  • Once I put my sampling method in action it become clear to me that it’s not as accurate as I want or that it needs to be. I could not use the 1mx1m frame at every quadrat like I planned. I can’t use the frame for the accessible quadrats and then alter my strategy by eyeballing a 1mx1m quadrat for the inaccessible ones.
  • I need to review different sampling methods, particularly options of sampling from a distance.
  • Finding the single juvenile Alnus rubra brought more variables to mind such as; how exactly am I defining what a juvenile Alnus rubra is and do I need to factor in the harsh exposed conditions that may slow their growth so significantly that what looks juvenile is actually much older?, could the red alder I find be from seed washed up from somewhere else?, how can I 100% determine if alders near a mature specimen are not vegetative growth?

How will modifying my strategy improve the research?

  • Clearly identifying what a true individual alder with defined juvenile characteristics is will reduce the chance of misleading or recording false data.
  • If I can find another sampling strategy that I can do from a distance with a monocular (once the tide is out!)I can have access to more shoreline and therefore more sample plots and improve the accuracy of my data.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

I chose to do a Systematic sample by distance, and my sampling strategy worked quite well for collecting data. I had 3 transect lines along 3 different gradients 20 feet apart. Each transect was 50 feet long, with a point (one unit) selected every 10 feet to make up 5 points on each transect. Each point was a stake in the ground where I would measure the 7 closest vascular plants. The forest was quite accessible this time of year, since a lot of the plants had lost their leaves, it wasn’t too overgrown to walk through. It became problematic when I was trying to measure distance between some plants, especially in dense areas. I initially was going to take down data for 5 different plants at each point, but found 7 to be more useful due to the closeness of some plants, and more data to work with. Another difficulty I came across was the fact that a lot of the plants had lost leaves and made it tough to identify. I took a lot of images and spent quite a bit of time making sure they were the correct species.

The data was only surprising in that there were species here I did not know about. I knew ferns would be quite dominant, but didn’t realize that they would be prominent in a large portion of the selected points.

One thing I should consider, is that it may be of my interest to measure the density or circumference of each plant to get a greater understanding of how well they grow in each gradient. This would mean I would have to re-do the data collection, but I am not against the idea of narrowing down my study to a couple species. This might help focus my study and allow me narrow it down.

Post 5: Design Reflections

TRU Open Learning Leave a comment

I found my sampling strategy difficult to implement in that the samples could be clumped close together. In order to eliminate this difficulty I could try adjusting the number of steps associated with the random number generated, for example, only taking 10+ steps from my starting points instead of 9 or less.

I also found the data gathered to be difficult to work with. Douglas Fir circumference and ambient temperature do not seem to provide much information on their own in such a short timeframe. In order to make this data easier to interpret over the duration of this study I would like to add quadrant aspect and the relative subject population within each quadrant.

 

Post 5 – Design Reflections

TRU Open Learning Leave a comment

My initial data collection was not difficult.  The plot locations were easy to get to other than the 20cm of snow that was unpleasant, but otherwise the terrain is very accessible.  The data collected showed that there were actually very few regenerating cottonwood stems, which at first glance, I thought there would be substantially more.  There were also less coniferous species found than expected, but this could also be due to the fact that small stems are under the snow, or have been browsed.  The amount of woody shrub cover can make plots difficult, as there are a lot of plants to maneuver around.   The only modification I may make is increasing plot size from a 3.99 meter radius (50 meters squared) to a 5.64 meter radius (100 meters squared).  This could improve the accuracy of the estimated density of tree species, as a larger plot may pick up more species, improving the estimated density.  It may also be helpful to actually count the number of shrub species in the plot as opposed to estimating their percent (%) cover, as these plants could be the main reason as to why there are very few coniferous species establishing in the area.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Did you have any difficulties in implementing your sampling strategy?

Yes, ~6 trees I was not able to record data for as the photos I initially took to analyse were too dark or too bright to tell the relative colour of the leaves. If I had done this analysis earlier I would have enough time to revisit those 6 trees to make a larger potential data set.

There were some surprises in the data. There was 1 tree in the thicker part of the forest that had 90% healthy leaves, which was much higher than the other trees in the same area (0-59%). You would expect that the area with the thickest forest would have hurt the growing chances of the tree. In this case the tree was quite tall in that. It was above most of the other tall trees, so this makes sense. It must have started growing before the younger trees had caught up to it. This area was recently logged so it gave at least a couple arbutus trees a chance to grow tall enough before the original forest took over.

There was one other sample that was interesting. It was Arbutus #7. It was a relatively small arbutus in an open area with few neighbours. 1 tree was directly overhead but wasn’t large, so maybe in this case the main issue wasn’t shade from the other trees but direct composition in the soil with neighbouring trees and a larger amount of brambles and bushes.

Otherwise the data did line up with expectations. Almost universally the the forested Arbutus were not healthy and small and larger healthier arbutus were found in the open areas and cliffs with fewer tall surrounding tree neighbours (including one huge tree with over 1000 leaves and with virtually zero damaged leaves).