Category: Post 5: Design Reflections

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Blog Post 5: Design Reflections

 

My initial data collection took place July 14/19 at 13:00 hours. I used 5 replicates of 1m2 quadrats with locations chosen using a simple random scheme. The response variable was % coverage of species within the quadrat and the predictor variable is elevation along the slope. To determine the % coverage of each species within each quadrat, area was calculated using measurements from a tape measure, compared to the entire area of the quadrat, and converted to a percentage. The only difficulty I found with this method was that it can sometimes be difficult to get accurate measurements for the irregular shape of plants. My data indicated that less complex plant species (ferns and clover) were more abundant near the base of the slope, and I began to find more complex species (Saskatoon berry) at higher elevations. I was not surprised by these results. I predicted that complexity of species along the elevation gradient would increase. This was because I figured that the hedges on the opposite side of the field would block sunlight from reaching plants at lower elevations, leaving them without enough access to the resource, resulting in stunted development. I will continue to use this method of data collection as I determine % coverage at higher elevations.

Blog Post 5: Design Reflections

TRU Open Learning 1 Comment

I used the area random sampling method. I built a 0.25m2 quadrat to determine bud density and measured the width of base branch growth for each replicate to gain an understanding of the effects of crowding on white spruces.The main difficulty that I encountered in my collection of data was in Location 1, the trees are so close together it is hard to walk between them. It was very hard to find my tagged replicates and carry my quadrat, measuring tape and field journal while fighting through the dense branches. Sometimes the outer branches of trees overlapped those beside them, making it harder for me to distinguish which buds belonged to the replicate I was sampling. The data was not surprising, it aligned with my hypothesis, which is that trees that are subject to crowding will be less productive than those that have ample space to themselves. The new growth bud density in Location 1 (most crowded) was on average lower than both that of Location 2 and 3 (least crowded). Despite my difficulties, I think this sampling strategy is the best one for my project. I still have to figure out how I am going to test the soil properties for each location (if I am even able to do so).

Blog Post #5 – Design Reflections

TRU Open Learning Leave a comment

This weekend I returned to my study sight to test out the data collection method I’d designed (outlined in Blog Post #3).  Saturday was a bust due to poor weather conditions, but Sunday afternoon looked a lot better.  I brought along the data collection tables I’d designed with the plan of collecting data on 3 individuals from each of my 4 species (Cormorant, Canada Goose, Franklin’s Gull and Mallard) for a total of 12 birds.  For each bird, I recorded their behavior at 15 second intervals for a total of 5 minutes, noting the location of each behavior along my gradient (Shore→ Shallows → open water).

Replicate: individual birds

Response variables: behaviors (categorical)

Predictor variables: species (categorical), time of day (categorical: AM/Midday/PM), point on gradient (categorical)

Panoramic view of the large pond

A few limitations and problems I noticed when I got to my site and started collecting data:

  • I hadn’t planned HOW I was going to select individuals to study in order to avoid bias.  Naturally, I was drawn to the most active birds who would be interesting to watch for 5 minute intervals.  I was also drawn to the birds closets to my location on the pond.
  • I realized that my lofty goal of trying to record the behaviours of multiple individuals from 4 different species over 3 different daily time periods might have been a bit over-enthusiastic for this project. The Franklin’s Gulls, for example, DO NOT HOLD STILL!  This species was frequently in flight, touching down for only brief periods.  The range of their flight paths made it impossible to ensure I was watching the same individual over the course of 5 minutes.
  • I realized that the pond is actually quite a bit bigger than I realized when I needed to identify a Mallard from other similar looking duck species from a distance.
  • My observations led me on a full loop around the pond, stopping to collect data when I saw birds of interest.  Again, this isn’t a very standardized procedure and could lead to bias when large groups catch my eye.
  • 3x 5 minutes of behaviour observation is not a very significant period of time over the course of a 24 hour day. Will  this be truly reflective of behaviour patterns?
  • The larger birds (Cormorants, Canada Goose) seemed to each have claimed specific territory around the pond.  There were no observation sites that allowed me to view both species at the same time.
Sample data collection table for the 4 species of birds observed

Reflecting on my trial run this weekend, I’ve come up with a few modifications to my research project:

  1. I plan to keep using the data tables I created as I found them easy to use and well laid out for the data I was collecting.
  2. I’m going to narrow my focus from 4 species to 1, the Mallard.  This species was found at many locations around the pond, and at all points along my gradient.  They were present in the highest numbers as well, giving me plenty of subjects to sample from.
  3. I’m going to use a randomized number generator (ie: 1-10)  to select my subjects: I’ll count to the random number, starting from left to right across the pond, and collect data on that individual. This should eliminate bias in choosing subjects.
  4. I’m going to select one observation point to work from, in order to prevent bias from wandering around looking for birds.
  5. Now that I’m going to be observing 1 species instead of 4, I will increase my number of subjects sampled each visit from 3 to 5, and increase my observation time for each individual from 5 minutes to 10 minutes. Doubling my observation time should provide slightly better behavior data.
  6. I’ve ordered a pair of binoculars off Amazon Prime, they’ll be here Wednesday!  This should help me identify Mallards from other similar looking ducks and allow me to record data across the pond from a fixed location.

 

It appears Team Canada Goose has also claimed this bench for themselves…

Based on these modifications, my hypothesis requires some adjustment as well.  I will keep the hypothesis that the water bird species studied will display increased levels of higher-energy activities (flight, feeding, etc) in dusk/dawn periods due to cooler temperatures, and increased display of lower energy activities (comfort, resting) mid-day when temperatures are higher.

Again, the null hypothesis would be that time of day has no effect on the time-activity budgets of water bird species.

Based on my research on Mallards thus far, I also suspect that typical behavior patterns will vary across my gradient, with resting/comfort behaviours being observed on land, feeding in the shallows, and locomotion/alert behavior taking place in open water. Mallards are considered “dabbling” ducks and feed by grazing on underwater plants indicating that I predict that I will see these behaviours most often in the portion of the gradient I have designated at “Shallows” (< 5 m from shore or visible plant matter appearing on/near the surface)

 

A view of the algae cover near the edges of the west side of the pond

Post 5

TRU Open Learning Leave a comment

I used the haphazard / area sampling strategy.  The only difficulty I had was finding enough replicates.  I had to return to the site and establish two more plots to ensure that I had 10 replicates for each environmental gradient. I choose the haphazard strategy because I felt if I used random or systematic plot location that I may not have found enough representative Cedar trees within a reasonable amount of plot locations.  My only concern with haphazard is the potential for introducing bias into the research.

I will continue to use the established plots but will begin collecting additional information specifically for each tree.  Every Cedar tree within my plots has been marked and numbered.

To better test my hypothesis and if had the time to set up a long-term research project I think I would design this experiment to be manipulative in a controlled environment.  I think a manipulative experiment could help determine the level of resistance with better accuracy.

Post 5: Design Reflections

TRU Open Learning 3 Comments

For the initial collection of data I used what I initially thought was a systematic sampling technique because I walked a certain number of paces to get to certain areas of each corner of the field. However, I realised that I was not actually using a systematic sampling technique because the sample taken in the middle of the field was not able to be measured by a certain number of paces equally from each “edge” of the field. This was due to the field not being uniform. Hence, I determined that I used a simple random sampling technique for the initial collection of data. This was not so much a difficulty but more of a misunderstanding on my part for sampling.

My initial hypothesis stated that there would be more dandelions towards the northern and eastern perimeters of the field because I had typically observed less prolonged human activity in those areas in the field whereas the southern and western parts of the field had benches, a playground, and a street hockey area. The data from the replicate in the southwest corner of the field surprised me because it had the greatest number of dandelions in the quadrat I set. However, I did not initially consider whether I would count flattened (versus upright) dandelions in my counts.

I still intend to use the random sampling technique for future data collection although I may consider adding more replicates to get a better idea of the abundance of dandelions in different parts of the field. One thing I may have to consider is that the park is maintained every so often and the dandelions may be mowed down during subsequent samplings.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Field Research Project

Posted on May 17, 2019 by caudia

Location: Courtenay Estuary/ K’Omosks Estuary

Date and time of site visit: April 21, 14:20

Collected by:  Cathy Audia

Before traveling to the estuary I check the tide schedule to avoid the problem of arriving and finding the study site submerged in water. Luckily this initial step, allowed me to collect the data with no other problems arising. While collecting the data, I was surprised to find the quadrants were not an accurate representation of the distribution of the subject. As the estuary is so vast I felt I needed to expand the number of quadrants recorded to allow for a larger service area to be represented.  This modification should allow for a more accurate representation of the subjects cover, resulting in more precise findings of the field study.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

I did not have too many troubles collecting data, I would like to continue though so that I have a larger number of days of data and my numbers will be more reliable. I am going to change the predictors of rain/no rain/length of rain to exclude length of rain- this is going to be difficult to quantify. I have also decided to expand my gradient to define whether it is sunny/partially sunny/cloudy where partially sunny means ~50% sun/cloud. The rest of my methods have been working well and I will keep the same.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

Here is a description of my sampling method:

A simple, random and distance based sampling technique was used, incorporating plots along a transect and transects along a path.

Each transect consists of three plots, one immediately adjacent to the pathway, another 10 m away from the path, and another 20 m away from the path. Each location for a transect was selected by entering the range of the number of metres of the trail length, from 0 to 3100 metres, for the trail along which data was to be collected. The first 5 numbers were selected, representing the number of metres from the beginning of the trail at which a transect would be located.

Each proceeding transect alternated from being on the left for the first to the right for the second, etc. The second plot was found by moving in the exact same direction, 10m away from the first plot centre, further into the forest. The third plot was found by moving in the exact same direction, 10m away from the second plot centre, further into the forest.

Upon finding the location along the trail, plot centre was found by moving into the forest 3m to the left (3m is the plot radius ) so that the plot is touching but not overlapping with the path edge. Each 3m plot encompasses 28.27 square meters and was selected based on silviculture survey practices. At each centre a stake was driven into the ground and a plot cord with a mark 3m away from the stake, marked the edge. If the mark touched the bark, a tree or shrub was included in the count.

The first plot represented the site with the most disturbance, the second plot represented the site with an intermediate level of disturbance, and the third plot represented a site with the least disturbance.

Here is my reflection on the sampling method:

I did not have much trouble with my sampling strategy. This was largely due to experience having conducted silviculture surveys over the past summer. I was nervous before and during the first number of plots as I had my fingers crossed that my predictions weren’t not true and I believe the data verified that. To find distance markers, indicating the placement of transects and distances between plots, I used my cell phone’s GPS technology and this probably introduced inaccuracies in distances as it is not as accurate as using an actual GPS with a low degree of error.

Some of the data collected did not seem to make sense. For instance, the number of large trees was often greater closer to the path. Though, my basic hypothesis is that they would be more plentiful further away from the path. When looking at the trees, the largest ones with a circumference of over 1m were most often found furthest away whereas trees considered ‘large’, more than 2m, may have been more plentiful but had a much smaller circumference. With this adjustment to sampling, adding another ‘class’ of tree size, I was able to reconcile the data on the ground and my hypothesis.

Blog Post 5: Design Reflections

TRU Open Learning Leave a comment

The collection of my initial data for my research project did prove to be a little challenging and I quickly realized some of the mistakes that I made. I was using a point count sample method in my location to count bird presence with ambient temperature as the predictor variable. However the sampling area was too large and therefore not the most effective way to sample. I used a grassed backyard area around 24 feet x 30 feet as the location which proved to be too confusing as I didn’t know whether to include birds on the fence. Also with birds flying in and through the area I wasn’t sure if I was double counting them. Therefore sometimes I counted them and sometimes I didn’t as I wasn’t sure if I had already. It was difficult to know whether the birds I was counting were ones that had already been counted. In hindsight I should have used a bird feeder on one of the trees and counted bird activity at the feeder.
Secondly, I also realized that my hypothesis was not detailed enough. My focus initially was hypothesizing that bird activity would be increased with warmer spring temperatures above 12.5 degrees C but I should have used a temperature range of 10 – 15 degrees C to hypothesize that temperatures outside of these ranges would have decreased bird activity because I needed to include temperatures both above and below the range as bird activity may be diminished in extreme temperatures on either end. I also should have done my testing in the morning when temperatures were cooler but due to time constraints I tested in the afternoon when temperatures were warmer and therefore I mostly had temperatures above my hypothesis. In hindsight I should have tested early in the morning when the temperatures weren’t as hot.
The results that surprised me were that the bird activity was strongest on the coolest day. I had hypothesized that birds prefer warmer weather but in fact based on the initial results they seem to prefer cooler weather.
When I test again, I plan to adjust my point count method by using the bird feeder. This will make it easier to count the birds, eliminate confusion and I feel it will be a more effective method. I also plan to conduct the research in the morning when temperatures will be cooler and slightly more variable (before the temperatures are at the height of the day). This will ensure I get enough days both above, below and within the range of my hypothesis. I also need to clarify my hypothesis before I complete the research project as it is too vague and doesn’t account for hot weather.

Post 5: Design Reflections

TRU Open Learning 1 Comment

I had a few difficulties with my random stratified sampling strategy. I was using a random number generator to get coordinates on where my quadrats should be for sampling. It took a long time to find coordinates in each strata that worked. It was also difficult to find the coverage of moss in each quadrat because I was counting the individual squares. The average time for one quadrat was approximately 30 minutes.

I was surprised to find moss in areas in an open field (0% canopy cover). It was surprising because the area gets direct sunlight the whole day, and horses are present in the field. I was also surprised to find moss in areas with almost full canopy cover, considering the ground was either covered in leaf debris or conifer debris (needles).

I plan to modify my approach in data collection. I’m going to use a piece of paper which is 10 cm x 10 cm. This will make the paper be 100 square cm and will be handy in counting moss abundance in the quadrat. The paper could be folded in half to represent 50 squares, in a quarter to represent 25 squares. This technique will make collecting data much easier. If the moss doesn’t completely cover the paper, I can subtract the squares moss isn’t present. This modification will improve the accuracy of data as well as shorten the sampling time for each quadrat.