Using an offline ensemble Kalman filter to estimate ammonia emissions in the Netherlands

Abstract. Ammonia (NH₃) emissions from agricultural sources remain highly uncertain, yet accurate emission estimates are essential for assessing impacts on air quality, ecosystem nitrogen deposition, and secondary aerosol formation. In this study, we apply an offline Localised Ensemble Kalman Filter (LEKF), where offline refers to the application of the filter to pre-computed ensemble model output without feeding the analysis back into the running model, to constrain monthly NH₃ emissions over the Netherlands for 2022, using ensemble simulations from the LOTOS-EUROS chemical transport model as the dynamical background. Observations from two complementary data streams are assimilated: ground-based in situ NH₃ concentration measurements from the MAN and LML networks, and satellite column retrievals from CrIS and IASI. The inversion system accounts for observation error characterisation, regularisation to prevent overfitting, and the application of satellite averaging kernels to ensure consistent comparison between retrieved and simulated columns. System performance is evaluated through synthetic perturbation experiments before application to real observations, which reveal systematic discrepancies between the prior inventory and observed concentrations, particularly during summer months when the prescribed emission time profile underestimates ambient NH₃. Finally, we assess observation network optimisation by identifying measurement locations that would maximally improve emission constraints across the Netherlands, providing practical recommendations for future monitoring network design.