Weather events that produce extreme precipitation can cause severe flooding and winds that result in thousands of deaths and billions of dollars in damages annually in the United States. Extreme precipitation events cover a range of spatiotemporal scales from 10s to 10,000s of kilometers, including mesoscale convective systems, tropical cyclones, extratropical cyclones, and atmospheric rivers. The objectives of the work proposed here are to improve our understanding of the small- and large-scale processes that govern these events and improve our ability to predict and prepare for their changes under the influences of natural variability and human activities. We propose to compare two promising directions in the future development of the Energy Exascale Earth System Model (E3SM) to investigate the tradeoffs between resolving the convective scale processes that control the intensity versus large-scale processes that control the dynamic structure of extreme precipitation events that impact the United States.

The first approach is a conventionally parameterized version of E3SM with high-resolution (~25 km) (HR-E3SM), which captures synoptic to mesoscale dynamics with one resolved scale. The second approach is a superparameterized version of E3SM (SP-E3SM), in which kilometer-scale cloud-resolving models are embedded in a low-resolution (~100 km) exterior framework of E3SM to explicitly represent convective processes with a second resolved scale. Each method has different benefits for simulating some aspects of extreme precipitation. They have not previously been compared under the constraint of consistent computational expense, nor have they been systematically compared across the range of scales relevant to weather systems that impact the United States year-round.

We will assess the HR-E3SM, SP-E3SM, and standard low-resolution version of E3SM (LR-E3SM), with a combination of climatological simulations (AMIP) and short targeted ensemble hindcast simulations (CAPT) for each storm type, against long-term and field-campaign-based observations. Our analysis will address three main objectives: 1) Determine what types of extreme precipitation events (i.e., mesoscale convective systems, tropical cyclones, extratropical cyclones, and atmospheric rivers) contribute most to precipitation intensity biases over the United States in the LR-E3SM; 2) Determine what the critical scales to resolve are (convective intensity with SP-E3SM versus dynamic structure with HR-E3SM) in order to capture realistic precipitation intensity for each storm type; and 3) Determine the contribution of each storm type to climatological precipitation distributions, clarify the consequences of underrepresenting interactions between small- and large-scale processes, and establish quantitatively whether superparameterization or high-resolution more accurately represents extreme precipitation given equivalent computational resources.