User: | Open Learning Faculty Member:
In designing a sampling strategy, time was of the essence. I began the course in October when the vegetation had full foliage but I knew that within the next week or two, fall would begin, and the plants would begin to lose their foliage or die off for the winter. I was able to go out for a couple of successive days to observe the environment and decide on a topic and strategy. In deciding to investigate light levels with vegetation coverage, I realized that the most prudent approach would be to quickly obtain a photographic record of the study area and analyze the vegetation cover from the photographs rather than studying the plants in-situ. This approach utilizes computer data processing tools with methodologies widely used in Remote Sensing studies and utilizing recognized academic software, giving it credence for a scientific inquiry.
I selected the starting point of the path (where it branched off another path segment) and recorded data at approximately 10m intervals. Capturing the data and photos at each interval went quickly so rather than random sampling, I collected data samples along the entire transect and I can either use the complete data set or randomly select points along the transect and query the data collection at these points. Again, a measure of expediency was directing my work because I realized that if there was a data deficit, I would not be able to collect the data again during the course of this study.
I would have preferred to construct an apparatus to ensure consistent height and angles for the camera but I compensated for this by using a highly coloured measuring stick in each photo where the spatial integrity can be confirmed. Additionally, since the same colours are available in each photo, I can calibrate the images’ colour balance and luminosity in much the same manner as NASA used with the Mars rover images.
I would have preferred to use a laboratory-grade light meter whose calibration could be independently certified. However, this was not possible due to the cost of such equipment, so I calibrated my meter statistically to ensure that its tolerances were of an acceptable level for the study. Also, I am not actually interested in the specific Lux values from the meter, but rather the relative light readings from one station to another. As such, it was only to necessary to confirm the light meter’s response relatively across the spectrum and not to specific laboratory standards. In other words, if the same level of light gives the same results on separate readings, I am able to consider the instrument calibrated for the purposes of this study.
I did the image analysis later at my home using the ImageJ scientific image analysis software. Ideally, I would have liked to write a macro to bulk process and analyze the images automatically but the complexity of this task was too much for the scope of this project. Therefore, I used personal judgement in some of the image processing which unfortunately could introduce bias or error but I endeavoured to follow the level of precision in the analysis and I believe that any discrepancies in analyses are statistically insignificant and do not affect the overall analysis.
