Adjoint model enhanced plume reconstruction from tomographic remote sensing measurements

Abstract

Abstract A new mathematical optimization method is presented for reconstructing pollution plume concentrations from tomographic remote sensing measurements on neighborhood scales (about 1 km × 1 km) using Differential Optical Absorption Spectroscopy (DOAS). The new method, called CAT–4Dvar, combines Computer Aided Tomography (CAT) and 4D variational (4Dvar) data assimilation. The objective of the method is to produce accurate reconstructions compared to the Algebraic Reconstruction Technique (ART) and other non-variational methods with only a small number of DOAS telescopes. A forward and adjoint 3D grid dispersion model was developed based on advection and diffusion solvers commonly used in air quality modeling. The adjoint model optimizes the model emissions and horizontal diffusion coefficient based on the difference between tomographic DOAS observations and ray path-integrated concentrations predicted by the forward model. It also updates the corresponding error covariances based on the Hessian of the cost function. An enhanced reconstruction is obtained from the forward model with optimized parameter values. In a synthetic experiment involving two hypothetical DOAS instruments, the CAT–4Dvar method yielded excellent results compared to ART, reducing the overall nearness index from 57% to 11%.