Fan Wu - Research

NASA Health and Air Quality

This project uses NASA land surface data to improve air quality models used in state implementation plans, required when states fail to meet EPA pollution standards. States run numerical models that simulate weather conditions and atmospheric chemistry.

My work (yellow highlights) focuses on the meteorological inputs (see below for details) by 1) analyzing observational data to evaluate state agency's current modeling system, and 2) assimilating remote sensing data to improve model performance.

Research areas and collaborators:

California San Joaquin Valley (SJV): intense agriculture with irrigation
- California Air Resources Board (CARB)
Multi-state Mid-Atlantic region (MMA): mix of urban, rural and forested land
- Pennsylvania Department of Environmental Protection (PA DEP)

Project overview. Figure credit: Alex Zhang. Modified: Fan Wu.

Observations: Land Surface and Boundary Layer

I analyzed surface heat, moisture, and momentum fluxes using eddy-covariance (EC) data from over 30 AmeriFlux sites across the U.S., including rare irrigated agriculture in the SJV. I helped deploy two new EC towers in the SJV, data now publicly available (US-NSa; US-NSb).

I also used ground-based lidar ceilometers from CARB and PA DEP to extract atmospheric boundary layer (ABL) height and cloud fraction at four SJV and two PA sites—filling gaps where rawinsondes are unavailable.

To improve land surface inputs, I used NASA MODIS and SMAP satellite products (e.g., LAI, GVF, soil moisture) and a remote-sensing-based irrigation scheme for data assimilation in the model.

Model domains (red rectangle) for the two areas of interest, each at 4-km resolution.

Flux tower data QA/QC for sensible heat flux (others not shown) from the flux tower at US-NSb in the SJV.

Models: WRF and LIS Data Assimilation

State agencies like CARB and PA DEP use numerical weather prediction models, i.e. the Weather Research and Forecasting (WRF) Model to simulate regional weather and air quality.

I evaluated WRF and used NASA's Land Information System (LIS) to assimilate remote sensing data into the Noah-MP land surface model (LSM), and then coupled with WRF for ABL simulations (NU-WRF/LIS).

I designed seven model experiments with varying configurations to assess biases. My results show improved representation of irrigation and land use, leading to better surface energy flux simulations and potentially more accurate ABL development.

Reanalysis (black) and remote sensing (red) datasets used in NASA LIS with Noah-MP LSM.

Land use/land cover change in the SJV (top) and MMA (bottom), before (left) and after (right) incorporating remote sensing data.

Applications: Air Pollutant Dispersion

The development of the ABL directly affects the dispersion and transport of air pollutants. I compared observed ABL heights and cloud fractions—derived from ground-based lidar backscatter—with ABL simulations from WRF and NU-WRF.

We found that incorporating land surface remote sensing in the models led to more accurate surface fluxes and ABL characteristics—more accurate latent and sensible heat partitioning, lower PBL heights, reduced near-surface winds and temperatures, and higher relative humidity. These improvements reduced PM₂.₅ and ozone biases, highlighting the importance of data-informed surface processes in air quality modeling.

ABL dynamics that connect surface fluxes with ABL development and haze extends on a hot summer day.

ABL heights (black dots) derived from Lidar backscatter (colors) on July 7, 2021, in Sacramento, CA.