RETRO: City-Scale High-Resolution Footprint Modeling Using LPDMs

Surface flux footprints are used to link gas or aerosol emissions with atmospheric observations. These footprints quantify the spatially explicit source-receptor relationships between surface emissions and concentration measurements at a specific receptor location and time. Lagrangian Particle Dispersion Models (LPDMs), such as STILT, FLEXPART, or HYSPLIT, are widely used to compute these surface flux footprints. Most footprint-based inverse modeling studies optimize surface fluxes on a country-, continental-, or global scale. Our focus is on predicting surface emissions at a much finer scale, with horizontal resolutions as small as 100 m, using building-resolving meteorological fields with horizontal resolutions as small as 10 m.

 

RETRO (REgional TRansport Operators for atmospheric inverse modeling) is our newly developed atmospheric footprint tool targeting regionally constrained inverse modeling approaches on city-scale domains. RETRO uses HYSPLIT (in STILT mode) as well as MPTRAC (an LPDM) under the hood and introduces various refinements over the original STILT model. This presentation will highlight three of these refinements: background concentration estimation, surface-emission coupling, and ultra-high-resolution footprints (~ 10m).

 

First, we address how RETRO handles concentration variations coming from outside the modeling domain. This is necessary because the concentration observed at a specific location is the result of both nearby surface fluxes as well as a background concentration. We compare different approaches to estimate the spatially and temporally heterogeneous background concentration of urban sensor networks, using either observations (e.g., using an upwind site) or global/regional model products (e.g., CAMS, ICON-ART, WRF-CHEM). Furthermore, we discuss different approaches to connect surface emissions with the particles simulated by the LPDM. The original STILT formulation does not account for elevated emission sources and can be inaccurate for sources near the receptor. We compare existing solutions to these cases, as well as our new approach implemented in RETRO, to refine the STILT formulation. Lastly, we show how 10 m resolution building-resolving CFD wind fields from GRAMM/GRAL can be used to compute ultra-high-resolution footprints in urban areas using the LPDM MPTRAC.