SIParCS 2021 - Dianne Deauna
Easier, Better, Faster, Shorter: Updates to grid-aware analysis with xgcm
Ocean models with varying resolutions are powerful tools for analyzing and predicting ocean states over historical and even future time periods. These models divide the ocean into cubes or grid cells, where cell averages of ocean variables are calculated by integrating partial differential equations forwards through time. Models represent scalar (e.g., temperature) and vector (e.g., velocity) quantities on separate positions within a grid cell. This requires that post processing model outputs must account for model metrics (essentially the relationships among different positions within a grid cell, e.g., distance along the x-axis between two temperature points vs two u-velocity points). The xgcm Python package was developed with the ability to apply grid-aware operations such as averaging, integration, and differentiation (among others) across different models in a convenient and highly efficient manner. This talk will focus on updates developed over this summer for xgcm’s handling of grid cell geometries in GCMs, more specifically on assigning model metrics, guessing them when necessary by interpolating across different dimensions, and selecting the appropriate metric for a given operation. These updates have made xgcm more user-friendly, as demonstrated by an accessible, editable Jupyter notebook provided through Binder.
Mentors: Anderson Banihirwe, Julius Busecke (Columbia University), & Deepak A. Cherian
Slides and posters