by Tanvi A.
Photo: Lakeside students ascend Mount Baker on a 2019 outdoor trip.
The Lakeside Summer Research Institute (LSRI) is a four-week summer research experience in which students engage in mentored research projects.
My research project in LSRI 2021 was to compare temperature trends on the north and south sides of Mount Baker, during which I learned about the process of data munging (refining raw data) and working through errors in data science. Mount Baker is located in the Northern Cascade Mountain range and is one of the snowiest mountains, making for unique weather and temperature patterns. Because wind flow and geographic features differ on either side of the mountain, if temperatures are similar on both sides, forecasters will be able to generalize Mount Baker’s predictions. I used Lakeside’s iButton temperature sensors, deployed by students on an outdoor trip in 2019, for data on the south side and Northwest Avalanche Center (NWAC) sensors for data on the north side from July 2019 - 2020. To ensure accurate conclusions, I compared a north and south sensor at both a lower altitude and a higher altitude.
When tackling this research question, I spent most of my time troubleshooting errors when trying to format the various datasets in my Python program. Through this, I started to learn that science is not as much of a linear process as I thought. With every new problem that I resolved, there were always other parts of my program that I had to go back and change. Although the iterative nature of science was frustrating at times, I learned a lot more from this. Once I was able to get lists of daily average temperatures for each of the four datasets, I had to adjust the values to account for the slight altitude difference between the corresponding sensors. Because of missing data points in one dataset, this led to errors when dealing with a new data type called NaN (not-a-number). But, I ultimately gained a lot more experience in applying new programming methods. After this, I was able to graph a linear regression for both comparisons.
As I expected, the lower north vs. lower south comparison graph appeared to have a strong correlation with a slope of nearly 1.0 (see Figure 1). However, the upper north vs upper south comparison graph wasn’t nearly as simple to analyze (see Figure 2). Around 0 °C for the upper south sensor, there was a vertical cluster of data points that didn't fit with the line of regression. I realized that since the upper south sensor was buried underground, unlike the lower south sensor, it doesn’t record surface temperature accurately when there is snow over the soil. With the help of a graph created by Tanvi G. ‘22 on the presence of snow cover with altitude, I identified and eliminated data points in the specified time period. Finally, I ended up with a graph with a slope of 0.9, which is more similar to the slope of the lower altitude comparison (see Figure 3). I was able to conclude that temperatures on the North and South Side of Mount Baker do have a strong correlation. I hope that future LSRI students will be able to use these results to make new predictions and continue uncovering more of Mount Baker’s various climate patterns.
Figure 1: Scatterplot of the correlation between daily average temperatures of the lower north and lower south sensors.
Figure 2: Scatterplot of the correlation between daily average temperatures of the upper north and upper south sensors. The vertical cluster of data points around 0 °C is due to the presence of snow cover on top of the upper south sensor.
Figure 3: Revised scatterplot of the correlation between daily average temperatures of the upper north and upper south sensors. The data points present during snow cover were not included in this plot.