Using dichotomous satellite observations of dust emission to improve dust emission modelling for future climate predictions

Abstract

<p>Atmospheric mineral dust has a significant impact on many Earth’s systems, including the energy budget, with subsequent alterations to the climate. These impacts occur directly, by altering the radiative properties of the atmosphere, or indirectly, through changes in cloud formation and precipitation rates. Accordingly, dust emission processes, determined by dust emission models are required to accurately predict climate response in climate change scenarios. Dust emission models remain poorly constrained either by parameterisations or available data and measurements of dust in the atmosphere have long been used for their calibration. However, there is growing recognition that this calibration to atmospheric dust confounds the magnitude and frequency of emission from dust sources and hides potential weaknesses in dust emission model formulation. In the satellite era, dichotomous (presence=1 or absence=0) observations of dust emission point sources (DPS) provide a valuable inventory of regional dust emission. We used these DPS data to evaluate dust emission model performance using coincidence of simulated and observed dust emission (or lack of emission). We evaluated the recently developed albedo-based dust emission model (AEM), with some inherited constraints it dynamically represents land surface roughness (vegetation to grain scale), varying over space (500m resolution) and through time (daily), improving predictions of wind friction velocity at the land surface ().  Using a total of 37,352 unique DPS locations aggregated into 1,945 1° grid boxes to harmonise data across the studies we identified a total of 59,688 dust emissions. The DPS data alone revealed that dust emission rarely recurs at the same location, even in North Africa and the Middle East (occurring 1.8% of the time), indicating that dust emission is an extreme, large wind speed event. The AEM over-estimated the occurrence of dust emission by up to 2 orders of magnitude, coincided with dichotomous observations 71% of the time but incorrectly simulated dust emission 27%. Our analysis indicates that a key constraint to dust emission modelling is that entrainment threshold is typically too small, needed to vary over space and time and at a scale consistent with the model. During observed dust emission,   was often too small because modelled wind speeds (ERA5; 11 km) were too small. The absence of any limit to sediment supply caused the AEM to simulate dust emission whenever <em>P</em>(<em>u</em><em><sub>s*</sub></em> > <em>u</em><em><sub>*ts</sub></em>), producing many false positives when and where wind speeds were frequently large. These results demonstrate the wind range of constraints to existing dust emission modelling, which will continue to hinder dust-climate projections. The improvement of dust emission modelling is essential against dust emission point source data to provide a consistent, reproducible, and valid framework for routine evaluation and potential model optimisation.</p><p> </p>