Estimating Tracer Sources with Ensemble Data Assimilation: Tests in an Augmented Lorenz-96 Model with Tracer Advection

Characterizing the source and behavior of airborne contaminants is an important problem in air-quality analysis and fighting air pollution. Identifying the location and strength of the source of a potentially harmful pollutant is often necessary to take appropriate actions for mitigation. At the very least, tracer releases from such sources need to be modeled to predict the damages they might cause. However, given the chaotic nature of atmospheric circulation, modeling airborne tracers with accuracy is a challenging task. To explore this issue, a novel low-order dynamical system is implemented to investigate the capabilities of ensemble data assimilation to estimate tracer concentration and sources. Our study coupled the low-order Lorenz-96 model with a Semi-Lagrangian scheme to advect model tracers on a one-dimensional periodic domain. We then assimilated high quality synthetic observations of a truth trajectory using the ensemble adjustment Kalman Filter (EAKF) inside the Data Assimilation Research Testbed (DART). We were able to assimilate observations of Lorenz-96 state variables (here serving as a wind) and tracer concentrations. Assimilating only wind observations improved analyses for both wind and tracer concentrations, but more interestingly, assimilating only tracer concentration observations noticeably improved the analyses for wind. Assimilating both wind and tracer concentration observations resulted in the overall best estimates. Assimilation of wind and tracer concentration observations also enabled estimates of tracer source location and strength. We explored the impact of gradually lowering the quality and density of observations on estimates of source location and strength and also the best locations for observations to detect sources at given model grid points.