My research group looks at the dynamics and predictability of weather extremes, and the potential of enhancing our current capability to accurately predict them using data assimilation techniques.
All-Sky Satellite Data Assimilation
Satellite observations possess the largest portion of assimilated observations and collectively the biggest impact on forecast accuracy in operational weather forecast models. Unfortunately, these observations also remain under-utilized, especially when they are contaminated by clouds and precipitation. We endeavor to adopt the “all-sky” approach that assimilates satellite observations from both clear-sky and cloudy/precipitating conditions. We primarily focus on the infrared radiances from geostationary satellites (e.g., GOES-East, Hiamawari-8, etc.) and microwave radiances from low-Earth-orbiting satellites (e.g., TRMM, GPM, Suomi NPP, etc.) and their data assimilation applications on tropical cyclones and severe storms. We have published over 10 peer-reviewed publications during the past 6 years showing great potential for these readily available yet underutilized observations to improve our forecast capabilities of tropical cyclones and severe storms, as well as their hazards.
The analyses of tropical cyclones and severe storms after assimilating satellite observations can be used to initialize forecasts and examine dynamical processes that govern the development of these systems as well as the hazards that are associated with them. We are also working on exploring more advanced approaches so that the all-sky satellite observations can be assimilated more efficiently and effectively to improve the accuracy of severe weather forecasts.
A high-resolution simulation of Hurricane Harvey using data assimilation analysis as input
Mesoscale Predictability
Predictability describes the longest extent to which we can make accurate forecasts. It is primarily constrained by errors in the initial conditions that are used to initialize the forecasts, errors in the models that are used to execute the forecasts, as well as the internal dynamics of the weather systems that modulate how errors grow. We have been continuously working on how different aspects contribute to the evolution of forecast errors, especially for weather systems at the mesoscale. We found that tiny initial uncertainties that are orders of magnitudes smaller than current global model analysis accuracy can amplify and propagate into progressively larger scales with the help of moist convective processes, gravity waves, and geostrophic adjustment. Therefore, there is an intrinsic limit of 6 to 12 hours for severe storms that they can not be accurately forecasted beyond this limit, and this limit will likely not be changed even with the progress of our observation systems and numerical weather prediction models.
Two simulations of severe storms that only have a difference of 0.1 K but lead to very different outcomes
Dual-Polarization Radar Observations of the Convective Boundary Layer
Most observations of the convective boundary layer nowadays come from the rawinsondes, but they are only available twice per day. Other remote-sensing profiling instruments can provide high-temporal-frequency observations of the convective boundary layer, but they are only available at limited locations. In collaboration with Drs. David Stensrud and Matthew Kumjian, we have been working on the potential of using dual-polarization differential reflectivity (ZDR) observations from the operational NexRAD network to characterize the depth of the convective boundary layer. Bragg scattering at the entrainment zone at the top of the convective boundary layer produces unique ZDR signatures that are relatively lower than the adjacent area. By converting radar range into altitude using the quasi-vertical profile approach, the entrainment zone at the top of the convective boundary layer can be represented by a “channel” or relatively low ZDR values.
Plan position indicator and quasi-vertical profile of ZDR from State College, PA, NexRAD radar (KCCX) on August 3, 2022
This is the first time that a nationwide convective boundary layer observation network with a much finer temporal resolution than the rawinsondes (every ~10 minutes) becomes possible. We are currently using these observations to understand the characteristics of convective boundary layer evolutions, how soil, land, and atmosphere processes contribute to their evolutions, what errors and biases the operational numerical weather prediction models have in predicting the convective boundary layer evolution, and the assimilation of these observations to improve the forecasts of severe weather events.