Chengsi Liu
University of Oklahoma, OK·Earth & Environmental Science
Short-range forecasts of severe storms depend heavily on data assimilation. This is the process by which a computer weather model combines its own forecast with real observations, such as weather radar scans, to estimate the current state of the atmosphere.
- Fields
Earth & Environmental Science
Computer Science
Physics
- Topics
data assimilation
machine learning
weather radar
- Grants
- 1 known grant, through Aug 2029
What they work on
- Develop a machine-learning-guided adaptive inflation method that corrects under-dispersive ensembles (forecast sets that are too similar to one another) in storm regions during radar data assimilation.
- Train machine learning models to learn bias fields that depend on the weather situation and on each variable, using radar innovations (observation-minus-forecast differences) and ensemble background forecasts.
- Test whether the method restores realistic ensemble spread while preserving balance between variables, and whether it improves analyses and short-range forecasts of severe storms.
More about the research
Better storm forecasts have major effects on public safety and the economy. Current methods for assimilating radar data have a weakness in how they balance trust in the model against trust in the observations. Forecasters often run many slightly different versions of the model at once, called an ensemble. In storm regions these ensembles tend to be too similar to one another, a problem called under-dispersion. As a result, the system gives radar data too little weight.
Why it matters
Better use of radar observations in weather models could improve short-range forecasts of thunderstorms and other severe weather. More accurate warnings could help forecasters, emergency managers and communities protect lives and reduce economic losses. The approach is designed to be computationally efficient, which matters for operational forecasting centers.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Steven Greybush
Pennsylvania State University, PA·Earth & Environmental Science
Design and optimize a deep learning nowcasting system for convection initiation that combines forecasting, diagnosis and generation of storm scenarios, trained on high-resolution satellite, radar and surface weather data.
Michael French
Stony Brook University, NY·Earth & Environmental Science
Establish a firm link between updraft area at middle and low levels and the occurrence of tornadoes and large hail in more than 100 supercells, using an automated algorithm that detects ZDR columns in dual-polarization radar data.
James Ruppert
University of Oklahoma, OK·Earth & Environmental Science
Determine whether real-world tropical Atlantic storm clusters self-organize or are forced by the large-scale environment, using SEA-POL C-band polarimetric radar aboard the R/V Meteor during PICCOLO.
The school
University of Oklahoma: 11 labs on PI Pages, $474M a year on research, 225 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.