Numerical Weather Prediction | Data Assimilation | HPC | AI Weather Models

I work on research and development around atmospheric modelling, numerical weather prediction, data assimilation, high-performance computing, and AI-based weather forecasting.

I developed and operationally integrated a dust-enabled MPAS-Atmosphere 8.0 forecasting system together with its complete HPC, conversion and post-processing workflow.

Major developments include:

• substantial redevelopment of the MPAS dust capability for the MPAS 8.0 code structure;
• native handling of the dust erodibility field (erod) in init_atmosphere;
• operational GOCART dust and aerosol configuration;
• native MPAS-to-WRF global remapping using an extended metgrid;
• generalized support for native MPAS atmospheric fields;
• Parallel-NetCDF support in metgrid through io_form = 11 and IO_PNETCDF;
• an extended UPP 4.1 workflow for processing the resulting global WRF-format fields;
• global GRIB production for GrADS, NCL and downstream operational processing;
• GeoJSON, polygon, isoband and MBTiles generation for operational visualization.

The global interoperability chain is:

Native MPAS forecast
        |
        v
MPAS-aware modified metgrid
        |
        | Parallel-NetCDF / io_form=11
        v
Global WRF lat/lon representation
        |
        v
Extended UPP 4.1
        |
        v
Global GRIB / operational products

View the operational MPAS system

I successfully developed and operationally integrated MPAS-JEDI/SABER background-error covariance workflows derived directly from native MPAS-Atmosphere forecast data.

The methodology has been implemented at:

• 12 km — successfully generated and operationally used;
• 24 km — successfully generated;
• 30 km — successfully generated.

The workflows include:

• VBAL;
• HDIAGS;
• NICAS;
• BUMP localization.

The covariance-training data originate from native MPAS forecasts, without requiring conversion from another numerical weather prediction model.

View the MPAS-JEDI tools and B-matrix workflows

• MPAS model workflows and utilities
• MPAS-JEDI data assimilation
• SABER / BUMP / NICAS
• Mesh conversion and remapping
• State conversion and diagnostics

• Operational WRF workflows
• WRFDA and FGAT
• HPC optimization
• Parallel I/O and domain decomposition
• Forecast post-processing utilities

• GSI data assimilation workflows
• Observation processing
• Diagnostic tools
• Radiance assimilation utilities

• ECMWF AIFS
• GraphCast
• PanguWeather
• NeuralGCM
• AI model integration and evaluation

• HPE Cray systems
• MPI and Slurm
• Parallel I/O
• Numerical model performance optimization
• Large-scale operational forecasting workflows

My current development work focuses primarily on MPAS, MPAS-JEDI, data assimilation, high-performance numerical weather prediction, and the integration of AI weather models into operational forecasting environments.

LinkedIn