In the atmosphere, the winds transport a large number of trace constituent species, such as water vapor, trace gases, and many types of aerosols, and these affect the climate. Therefore, tracer transport is an important component of global climate models. It is also computationally expensive, often the dominant cost in an atmospheric dynamical core. In models, tracers must satisfy several properties. This property preservation problem is the key challenge to overcome to speed up tracer transport.

We have developed methods and software implementations, called Communication-Efficient Density Reconstructors (CEDR), to solve this property preservation problem efficiently. These methods provide strong guarantees on computational and communication efficiency and tight bounds on the adjustments to a tracer field needed to preserve properties. They can be coupled to a wide variety of advection schemes. Because CEDR methods solve the key challenge in tracer transport, they enable speeding up tracer transport in the Energy Exascale Earth System Model (E3SM) Atmosphere Model (EAM), the focus of a larger project of which this work is one part. A faster model enables answering scientific questions more comprehensively.