Published January 16, 2025
By DeeDee DuPlessis
In the western United States, water supply and availability are closely tied to snowfall and the seasonal snowpack. The University of Wyoming’s project WyACT looks at how snowfall in this mountainous area might be affected by changing climate conditions. To forecast future scenarios, scientists rely on sophisticated models that calculate atmospheric and geographical conditions into temperature, precipitation, and other climate variables.
One such model is named CONUS404 because it covers the contiguous United States over 40+ years at a 4-km (2.5 miles) resolution. It is a historical climate reconstruction from 1980-2022 and was recently published in 2023. WyACT is using the CONUS404 dataset for various purposes, including to drive streamflow modelling for specific watersheds, and the project team is very interested in the quality of the model.
Kaitlin Smith, a master’s student in Atmospheric Science, took on the task of evaluating the model for her thesis. Smith, who holds a bachelor’s degree in Earth and Atmospheric Science from Cornell University, became aware of UW due to its proximity to the NCAR-Wyoming Supercomputing Center in Cheyenne, which is dedicated to research in Earth system sciences. Smith enjoys the interdisciplinarity of the work, which she notes is not always typical of an atmospheric science degree.
Kaitlin Smith assisting a balloon launch demonstration for the Global Warming undergraduate class
Comparing apples and oranges?
Smith compared the CONUS404 data against various other climate models as well as actual observations. In line with WyACT’s focus on snow and water, she concentrated on the amount and timing of precipitation, as well as the evolution of snowpack over a year. This task requires a significant amount of computation. For example, different models use different grid structures along which data points are distributed, and Smith needed to standardize them to make meaningful comparisons. The scale of the model, i.e. the distance from one grid point to another, can be a challenge in mountainous areas, where a few kilometers on the ground can mean many meters difference in elevation and therefore, temperature.
CONUS404 showed some discrepancies with existing data sets. Smith describes its behavior as “not intuitive”. Overall, cold-season precipitation and snowpacks during the early accumulation period aligned with other data, but the model showed the timing of maximum snowpack earlier and underestimates the depth of snow. Identifying the model’s strengths and weaknesses is a first step, but the basis for improving the model is to understand the reasons for its biases. This is complex, since the discrepancies could be due to the existing data as well as the new data set. Even observations are not necessarily accurate, depending on the exact location of the gauges. Models also tend to underestimate temperatures in the mountains.
Smith plans to continue working on the model and examine further questions. She is especially interested in the cycles of changes in snowpack across the day and night. She believes an analysis of the diurnal cycle could help to better understand the snow melt process. Fortunately, while working on her thesis, she made an effort to write reusable analysis code. Following the completion of her thesis, she is seamlessly sliding into PhD work under WyACT.
