Estimating methane fluxes in Northern Italy by inverse modeling: Evaluating optimal particle transport release heights in mountainous regions

The Po valley, situated in Northern Italy, is a flat region with mountains on the northern and southern ends, characterized by intensive animal farming and agriculture practices (including rice cultivation), all of which are significant sources of methane (CH₄) emissions. Although both sources and sinks of strong greenhouse gases are well identified, large uncertainties still remain in estimating the CH₄  budget. However, inverse modeling is an observation-based approach that can be used independently to verify existing emission inventories.
The objective is to estimate CH₄  surface-atmosphere fluxes in Northern Italy for the year 2019 over a nested grid covering the Po valley at 0.1°x0.1° horiz. grid res. using the atmospheric inversion framework FLexInvert+. The framework integrates atmospheric CH₄  mixing ratios from 17 ICOS sites, background mixing ratios using the CAMS inversion product, prior information (with total emissions of 594 Tg y^-1), and an atmospheric transport model to optimize surface fluxes to best match the observation. The FlexPart model is used to simulate the source-receptor relationship using ECMWF ERA5 windfields at 0.5°x0.5° horiz. res., and relates the surface fluxes to the changes in CH₄  mixing ratios. Modeling the dispersion of particles over complex terrain, i.e. the Alps, is challenging due to processes interacting with the orography, and a coarse model resolution smoothens the slope and elevation of the mountain. Hence, a sensitivity analysis was carried out for the six mountain sites >1000 m a.s.l. located within the nested domain to assess the optimal particle release height using 7-days backtrajectories. The inversion results from three different release heights were examined using 1) the original sampling height of CH₄ a.s.l., 2) the pressure-based height, determined by identifying the model-level height that minimized the difference between the modeled pressure and observed pressure at the receptor site, and 3) the potential temperature-based height, determined by matching the modeled potential temperature with observations to identify the model-level height. The hypsometric equation was applied to obtain the release height, and an averaged particle release height was selected for the entire year based on nighttime observations. Initial findings highlight significant differences in posterior estimates among the three release heights and hold promising prospects for achieving improved inversion results.