*By Swaraag Sistla*

I had only ever been to the University of Washington’s campus less than 5 times. Yet, I soon discovered within it lied an incredibly informative, engaging, and powerful experience that I will never forget. From building on my data science skills to finding a huge error in a public dataset to meeting students and professors at the UW, I truly grew a lot as a person and as a scientist.

My work this August centered primarily around analyzing the climate of Greenland. Using data science, I was able to produce numerous graphs that explained much about the various environmental variables of Summit Station, Greenland and how they relate to each other.

The graph above shows a one understanding for the four seasons of Greenland in four separate colors, with the y-axis representing the temperature two meters above the surface over time. We also noticed a fascinating correlation between the wind speed and time:

The summer months correspond to lower wind speeds, while the winters generally have higher speeds. We also calculated the inversion strength at Summit Station—the inversion strength is simply a measure of how much the temperature changes with respect to the height above the surface. In other words, for a temperature reading near the surface and for one high above the surface, the inversion strength tells us how much that temperature changes as the height above the surface increases.

Here, it is clear that the summer has very low inversion strengths, meaning the temperature does not increase with the height as much as it does in the winter months, where the temperature increases almost 1℃ per meter. We also created other graphs that grouped all values at a certain minute over the years and took their mean, resulting in a graph that shows the average temperature at some points in time over the years.

Perhaps our greatest discovery this month was related to the relative humidity at Summit Station, Greenland.

Expecting a graph for the Relative Humidity over 2 meter temperature that followed a straight linear correlation, we were surprised to see three distinct lines in the graph above. Confused as to what was causing such a strange set of three linear correlations, we graphed the relative humidity over time and saw three distinct segments of data that seemed to abruptly change into one another (highlighted as different colors in the graph below).

Believing miscalibrations could be affecting two of the three segments, we applied the Clausius-Clapeyron equation to determine which of these three segments is most likely the accurate one. After finding that the blue segment contained likely expected relative humidity values, we scaled up the other two segments to match the blue segment’s values using a Linear Regression approximation of the blue segment’s slope and y-intercept.