Earth System Models (ESMs) have notable biases on both regional and global spatial scales. Many of these biases are due, at least in part, to uncertainties in the parameterization of key subgrid-scale processes. One parameterizaton of interest is that of the planetary boundary layer (PBL). The Community Atmosphere Model version 6 (CAM6), the atmospheric component of the Community Earth System Model version 2 (CESM2), parameterizes PBL turbulence using the Cloud Layers Unified by Binormals (CLUBB) scheme. Past studies have demonstrated that uncertainties in modeled fields like cloud fraction can be attributed to CLUBB. My work further explores the sensitivity of CAM6-CLUBB output to certain tunable parameters in CLUBB, with an emphasis on a set of experimental input parameters that are not yet operational in CAM6. We perform a parameter sensitivity analysis that highlights several tunable CLUBB parameters that are influential for regional and global fields of surface wind stress (shown above) and cloud fraction. This sensitivity analysis serves as a launching point for exploring the physical mechanisms driving the evolution of parameterized turbulence in CLUBB, and how our understanding of these mechanisms can help model developers address established biases in CESM2.

Given the damage caused by tropical cyclones (TCs), it is critical that stakeholders receive timely and accurate TC forecasts at least several weeks in advance. In recent years, ESMs have been able to produce subseasonal to seasonal (S2S) TC forecasts with increasing skill. However, ESMs are still limited to relatively coarse horizontal resolutions for S2S forecasts, so small-scale processes need to be estimated, or parameterized. Studies have established that the parameterization of PBL turbulence is a key source of model uncertainty in forecasting TCs. My work explores the sensitivity of modeled TCs in CESM2 to the parameterization of the PBL. We find that the parameterization of turbulent mixing in the PBL affects the TC's structure, including the primary and secondary circulations and surface fluxes. We also show that these structural sensitivities can be leveraged in order to reduce errors in CESM2's depiction of TCs, which can then lead to improvements in S2S forecasting.

The public safety hazards associated with flooding, drought, heatwaves, hurricanes, tornadoes, etc. underscore the need to provide the public with sufficient advance warning to make reasonable preparations. Unfortunately, numerical weather prediction (NWP) models cannot reliably predict extreme weather more than several weeks in advance. Efforts to expand predictability of weather and climate extremes into S2S leads have focused on modes of intraseasonal climate variability like the Madden-Julian oscillation (MJO). Research has shown that tropical convection associated with the Madden-Julian oscillation excites an extratropical response (teleconnections) that drives North American weather several weeks to several months later. The MJO and its teleconnections are also influenced by the quasi-biennial oscillation (QBO), a ~2 year cycle of zonal wind anomalies in the equatorial stratosphere. My research has involved implementing an empirical model that predicts 3-6 week precipitation anomalies throughout North America using the historical response of precipitation to the phase of the MJO and QBO at initialization. Results showed that the model produces skillful precipitation forecasts 3-6 weeks in advance. I have also contributed to other studies that have demonstrated similar model utility for temperature and severe convection forecasts at S2S leads.

Atmospheric rivers (ARs) are long, plumelike corridors of high atmospheric water vapor transport, typically associated with extratropical cyclones. Upon interaction with complex coastal topography, ARs can produce significant precipitation along the coast and interior mountain ranges. For this reason, ARs are critical sources of seasonal precipitation, especially in the western U.S. Given the impacts of ARs, many studies have explored AR characteristics through the lens of both observations and models. To aid in these studies, various algorithms have been developed to objectively identify ARs in fields of water vapor transport. These algorithms typically employ criteria based on intensity and geometry. At Penn State, I actively develop and implement an existing AR detection algorithm in support of the Atmospheric River Tracking Method Intercomparison Project (ARTMIP). The goal of ARTMIP is to better understand the uncertainties in observational and modeling studies due to the choice of AR detection algorithm.

With increased computational capabilities, weather and climate models can now run with finer horizontal and vertical resolutions. These resolutions allow for a more realistic depiction of key physical mechanisms that govern the development of weather and climate. However, in spite of recent advances, weather and climate models still exhibit biases, namely due to uncertain initial conditions, spatial and temporal discretization error, and imperfect estimations of unresolved processes. I am particularly interested in the skill of models that predict weather and climate from the near-term to leads out to several months. For example, I performed an evaluation of short- and medium-range AR hindcasts from a variety of numerical weather prediction (NWP) systems. It was shown that after 10-14 days NWP models provide little additional skill compared to climatology-based forecasts of the geometry, intensity, and location of landfalling ARs along the U.S. West Coast. This erosion in the skill of NWP models after 2 weeks is consistent across a variety of precipitation-inducing weather features. Communication of these model limitations to end-users is crucial, so I am interested in developing novel techniques to further evaluate and communicate the benefiits and drawbacks of weather and climate models.