Constraining spatio-temporal variations in dust emission at global scale with ensemble data assimilation of satellite optical depth retrievals

Abstract

Mineral dust emissions play a fundamental role in the simulation of the dust cycle in numerical models. The emission of dust depends on a number of atmospheric and surface conditions that span a large range of time and spatial scales. Due to the inherent difficulties to physically represent this complexity in a simplified way, the emission of mineral dust is usually parameterized in the atmospheric numerical models. The heterogeneity of available dust emission parametrizations, along with the soil characteristics and meteorological information, the atmospheric models themselves, their tuning, and their boundary and initial conditions, contribute to the large spread of net dust flux estimated with different modeling frameworks.This work presents a novel approach to estimate dust emissions through the assimilation of dust optical depth filtered retrievals from satellite measurements, by means of an ensemble-based data assimilation scheme. Because of the lagged nature of the emission inversion problem, the assimilation is produced with a slightly modified version of the ensemble Kalman Filter algorithm. We show results of the inversion for 5-year global numerical experiments (2017 to 2021), by using dust-only simulations with three of the available state-of-the-art dust emission schemes implemented in the chemical MONARCH model.In these three experiments, we assimilate dust optical depth obtained from the SNPP-VIIRS Deep Blue retrievals. The control vector consists of model dust emissions at native spatial resolution (1.4 by 1 degrees) and a 3-days time resolution. We find regional and temporal corrections in the estimated emissions after assimilation that are consistent across the different dust emission scheme experiments, making our findings robust. We compare the dust optical depth of our simulations with the assimilated observations, as well as with independent dust-filtered optical depth from ground-based AERONET sun-photometers. The dust optical depth resulting from the simulations that use the corrected emissions show substantial improvements in the skill scores than the dust optical depth simulated with the uncorrected emissions. Our work paves the road toward quantifying and eventually reducing uncertainties in dust emission schemes and toward better constraining the contribution to climate of the dust sources at sub-regional scale.