Understanding the Influence of Parameter Value Uncertainty on Climate Model Output: Developing an Interactive Dashboard

Climate models are computer simulations that attempt to replicate the complex interactions between Earth’s systems. Improving the accuracy of climate models relies on evaluating uncertainty and minimizing error. The Climate and Global Dynamics Lab at the National Center for Atmospheric Research (NCAR) has recently carried out a Parameter Perturbation Experiment (PPE) to understand how the uncertainty of parameter values affected the output of their model, the Community Land Model (CLM); which simulates terrestrial processes. While the necessary data for the PPE has been collected, the data is stored in a collection of files that are difficult to interpret in their current form. The current website hosts visualizations for a portion of the PPE data, but contains no visualizations for data that more closely simulates Earth system interactions. These issues can be mitigated by developing an emulator with the internal complexity to isolate a one-to-one relationship between a parameter and climate variable, then display the predicted relationship. A publicly available emulator with these capabilities will allow scientists to easily interpret complex climate model outputs and offer insights on parameter-variable relationships that are not being predicted accurately by the model; which can lead to increased accuracy and precision of climate models. 

Acknowledgments

Bren School of Environmental Science & Management: Satie Airamé, Assistant Dean; Emily Case, Capstone Project Coordinator; Carmen Galaz García, Assistant Teaching Professor

Columbia University: Linnia Hawkins, Associate Research Scientist

National Center for Atmospheric Research: Nick Cote, Software Engineer; Daniel Kennedy, Project Scientist