Are We Getting the Right Answers for the Right Reasons Using Computational Models?
Colloquium's overarching questions are:
- Are we getting the right answers for the right reasons with computational models in Earth Science?
- Why should we care about being right for the right reasons?
What can we do to get the right answer for the right reason?
With the advent of new technologies such as sensors, geophysical surveys and remote sensing products, we are now amassing high volumes and wide variety of earth science, and critical zone, observational data over long periods of time. Modern bottom-up mechanistic models and top-down machine learning algorithms are becoming more and more efficient. What else do we need? The big unknown is: how much can we trust these bottom-up and top-down models? In this talk I show a few examples of hydro(geo)ology models (both mechanistic and machine learning) which are getting the right answers for wrong reasons. The potential implications of this artifact will be discussed for understanding the landscape evolution, landscape hydro(geo)chemistry response, watershed management, and drought and flood predictions. Finally, the opportunities to improve the reliability of mechanistic and machine learning models will be discussed.
Join Zoom Meeting
Meeting ID: 863 8927 4971