Statistical analysis of multi-annual time series of atmospheric mineral dust content in the Sahel.

Mineral dust has radiative and biogeochemical impacts, affects human health and soil fertility. The mineral dust cycle, i.e., dust emission, transport and deposition depends on meteorological parameters, in particular surface wind speed and precipitation. Climate change has lead to measurable change in surface temperature and precipitation regimes in the Sahel (e.g., Panthou et al., 2018) and is also expected to modify the surface winds that controls dust emissions and transport. 

Since 2006, mineral dust is monitored in the Sahel by the stations of the INDAAF network (https://indaaf.obs-mip.fr/). We used the PM₁₀ surface concentrations and the Aerosol Optical Depth (AOD) from the AERONET network measured in Cinzana (Mali) and Banizoumbou (Niger) to detect possible changes in the Sahelian atmospheric dust content. The Angstrom exponent is used to select situations where mineral dust is the dominant contributor to the AOD. PM₁₀ concentrations and AOD are significantly correlated but have distinct seasonal cycles, with a ratio PM₁₀/AOD peaking in August.

No clear trend on the annual and seasonal mean concentrations or AODs has been identified. When subtracting the mean seasonal cycle to the monthly median PM₁₀ concentration we observe a slight decrease of the residuals  in Cinzana (Mali) but no trend in the AOD. No correlation was found between the AOD or the PM₁₀ concentrations and the North Atlantic Oscillation Index but the PM₁₀ concentration tends to increase with the Sahelian drought index.  For most of the years, the PM₁₀ concentrations and AODs are lower when the maximum of the vegetation cover of the previous year (represented by satellite Normalized Vegetation Index) is higher. This may reflect the protective effect of the dry vegetation residues on dust emission. These results suggest that, for the measurement period (2006-2019), the variability of the dust content is mainly due to the seasonal cycle and that the year to year variability is so large that no trends can be detected. Longer time series, with a better temporal sampling, seem to be necessary to have a chance to detect a significant change.