by Varun I. ’22
"Twenty years from now you will be more disappointed by the things that you didn't do than by the ones you did do. So throw off the bowlines. Sail away from the safe harbor. Catch the trade winds in your sails. Explore. Dream. Discover." - Mark Twain
In my rising sophomore summer, I spent my time at the Lakeside Summer Research Institute (LSRI) understanding more about Seattle’s microclimates. In particular, I examined and analyzed differences between weather patterns from Sea-Tac and Lakeside. Inspired by the project’s success, this semester I pursued an independent study with Vishnu I. ’22 and Dr. Town where we aimed to utilize more advanced data techniques, like machine learning, to investigate weather phenomena on Mount Baker.
To kick off our project, Vishnu and I first read about the basics of machine learning — topics included regression, classification, and dimensionality reduction. Then, we wrote multiple programs to efficiently gather data from the Weather Research and Forecasting (WRF) model, a prediction system that provides forecasted data for meteorological variables in hour-long timesteps. Please read Vishnu’s blog to learn more about our learning and roadblocks during these weeks.
After we finished collecting the WRF data, we had arrived at the final stretch: visualizing our meteorological information and coming to justified conclusions. Pictured below is one of the graphs we produced, the incoming longwave radiation vs. air temperature (blue) and cloud overhead temperature (red), in January 2020 on Mount Baker (Fig. 1). Incoming longwave radiation, represented as W/m2, is defined as energy emitted from a cloud overhead or the atmosphere, which is then absorbed by the surface.
Immediately, one can notice that there is a trend in this relationship — for values of 240 to 260 W/m2, air and cloud temperature data appear to be tightly clustered with one another. To better understand why this is, we can use a modified version of the Stefan-Boltzmann law, Tc=(P/A * σ )1/4. In this equation, P/A represents the incoming longwave radiation, represents the Stefan-Boltzmann constant, and Tc represents the cloud overhead temperature. We see that when the incoming longwave radiation increases by a value, the cloud temperature increases by the fourth root of that number. Thus, in the interval between 240 and 260 W/m2, cloud temperature catches up to air temperature and the two variables have roughly the same value. In other words, the ground and clouds overhead are now in thermal equilibrium.
One lesson I learned in this project is that the knowledge of data techniques can only get one so far. Though I used classification and regression to better understand the behavior of data, it could not reveal the inner workings of the Earth. It could not tell me how much radiation was present at each stage of the atmosphere. All of our interpretations had to come through creativity and pondering about principles of geophysics.
Trying to accomplish this high-level research in four weeks wasn’t easy. It was certainly a leap of faith, but I took it. I only noticed success after I had been through failure — many graphs I produced didn’t have noticeable correlations, and some weather variables, like snow height, had corrupt data. But although I had sailed away from a safe harbor and battled my way upstream, in the end I arrived at an even safer one. A harbor that would now support me for the bigger war: using all of the lessons I learned to conquer complex datasets later on in life. To whomever is reading this, I encourage you to do the same: seek discomfort and throw off the bowlines of safety, because that is truly how you will dream, explore, and discover.
Figure 1: Shows the incoming long wave radiation (W/m2) vs. the air temperature (blue points) and cloud overhead temperature (red) on Mount Baker in Jan. 2020. Air temperature and incoming longwave radiation data were obtained from WRF files. Cloud overhead temperature was calculated using the equation Tc=(P/A * σ )1/4.