Praful Dodda
Ph.D. in Environmental Sciences and Engineering
Gillings School of Global Public Health, UNC-Chapel Hill
praful@unc.edu · Curriculum vitae · GitHub · Projects
I develop space-time statistical methods for estimating human exposure to air pollution, and the modeling infrastructure that supplies them with data.
My work centers on the Bayesian Maximum Entropy (BME) framework, which combines observations of differing quality (regulatory monitors, chemical transport model output, satellite retrievals, and low-cost sensors) into a single estimate with quantified uncertainty at every point in space and time. I have applied it to produce a 33-year global reanalysis of surface ozone at 0.1° resolution, to map exposure disparities to petrochemical air toxics across the U.S. Gulf States, and to evaluate whether low-cost sensor networks can be responsibly incorporated into regulatory-grade estimates.
Alongside the statistical work I build and operate the models that generate its inputs: chemistry and dispersion modeling for aircraft plumes under FAA sponsorship, emissions inventories from radar flight-track data, CMAQ sensitivity campaigns at the scale of a hundred simulations, and machine learning pipelines that translate satellite column retrievals into surface concentrations.
Research
Space-time data fusion
Bayesian Maximum Entropy and kriging methods for combining heterogeneous data sources, including nonlinear bias correction that converts model and satellite products into a corrected mean with a localized variance rather than treating them as exact.
Air quality and dispersion modeling
Chemical transport and dispersion modeling with CMAQ, WRF, AERMOD, and research aircraft plume models, including chemistry mechanism implementation, emissions processing, and formal model evaluation against monitoring networks.
Exposure and environmental justice
Translating concentration fields into population-weighted exposure metrics, quantifying how exposure burdens differ across demographic subgroups, and preparing datasets suited to health impact and burden of disease assessment.
Scientific computing
Machine learning on satellite retrievals, national-scale data pipelines, HPC campaign management and quality assurance, and geostatistical software built for use by researchers who are not its developers.
Selected work
- A 33-year global reanalysis of surface ozone. Monthly MDA8 estimates at 0.1° for 1990–2022, fusing TOAR-II observations with a bias-corrected CTM ensemble and multiple satellite retrievals.
- Exposure disparities to petrochemical air toxics in the U.S. Gulf States. Daily 4 km estimates of SBTEX and 1,3-butadiene showing a widening exposure gap between non-white and white populations from 2011 to 2016.
- Chemistry schemes in an aircraft dispersion model. Six NO2 chemistry treatments implemented and evaluated against AERMOD and explicit chemistry for the Los Angeles airport domain.
- MBMEGUI. Graphical software for Modern Bayesian Maximum Entropy space/time analysis, used in environmental science instruction at UNC.
Technical skills
Languages
- Python
- R
- MATLAB
- SQL
- Fortran
- bash / csh
- JavaScript
Models
- CMAQ / DDM-3D
- WRF
- AERMOD
- SMOKE
- AEDT
- MMIF / MCIP
- F0AM / MCM
Methods
- Bayesian Maximum Entropy
- Kriging
- Data fusion
- Bias correction
- Machine learning
- Risk assessment
Computing
- SLURM / HPC
- xarray & Dask
- scikit-learn
- TensorFlow
- PyTorch
- QGIS / ArcGIS
- Git
Contact
Department of Environmental Sciences and Engineering
University of North Carolina at Chapel Hill
praful@unc.edu