Performance of a dust model to predict the vertical mass concentration of an extreme Saharan dust event in the Iberian Peninsula: Comparison with continuous, elastic, polarization-sensitive lidars

Abstract

An intense dusty event unusually occurred in wintertime over the Iberian Peninsula was detected over two Spanish NASA/MPLNET sites: the temporary Torrejón Observational Tower for Environmental Monitoring (TOTEM, 40.5°N 3.5°W) and the Barcelona station (BCN, 41.4°N 2.1°E). The highest dust incidence was observed from 22 to 23 February 2017; this two-day dusty scenario is examined in order to evaluate the performance of the operational NMMB/BSC-Dust model on forecasted mass concentration profiling in comparison with polarized Micro-Pulse (P-MPL) mass estimates for dust particles. First, the optical properties of the dust (DD) were effectively separated from the non-dust (ND) component by using the combined P-MPL/POLIPHON method. Lidar-derived DD optical depths reached maximums of 1.6–1.7 (±0.1) at both stations. Typical features for dust were obtained: linear particle depolarization ratios between 0.3 and 0.4, and lidar ratios in the range of 41–70 sr and 36–66 sr, respectively, for TOTEM and BCN. Lower AERONET Ångström exponents were reported for TOTEM (0.12 ± 0.04) than at BCN (0.5 ± 0.3). HYSPLIT back-trajectory analysis showed air masses coming from the Sahara region, mostly transporting dust particles. AERONET-derived Mass Extinction Efficiencies (MEE) under dusty conditions were used for the extinction-to-mass conversion procedure as applied to the P-MPL measurements: MEE values were lower at TOTEM (0.57 ± 0.01 m2 g−1) than those found at BCN (0.87 ± 0.10 m2 g−1). Those results reveal that dust particles were predominantly larger at TOTEM than those observed at BCN, and a longer transport of dust particles from the Sahara sources to BCN could favour a higher gravitational settling of coarser particles before reaching BCN than TOTEM. A comparative analysis between profiles as obtained from the lidar DD component of the mass concentration and those forecasted by the NMMB/BSC-Dust model (25 available dusty profiles) was performed. The degree of agreement between both datasets was determined by the percentage of dusty cases satisfying selected model performance criteria (favourable cases) of two proxies: the Mean Fractional Bias, MFB, and the correlation coefficient, CC. A good agreement is found (72% and 76%, respectively, of favourable cases); however, large discrepancies are found at low altitudes between the dust model and the lidar observations, mostly at early stages of the arrival of the dust intrusion. Higher model-derived centre-of-mass (CoM) heights are found in 60% of the cases (with differences < 15% w.r.t. the lidar CoM, whose values ranged between 1.8 and 2.3 km height). In addition, modelled mass loading (ML) values were generally higher than the lidar-derived ones. However, the evolution of the mass loading along the two days, 22 and 23 February, was rather similar for both the model forecasting and lidar observations at both stations. The relative ML differences (<50%) of the mass loading represented 60% of all cases. Discrepancies can be based on the uncertainties in the lidar retrievals (mainly, the use of single extinction-to-mass conversion factors). In general, a moderately good agreement is observed between the P-MPL-derived dust mass concentration profiles and the NMMB/BSC-Dust model ones at both sites; large discrepancies are found at lower altitudes, plausibly due to a lower sedimentation of dust particles coming from upper layers by gravitational settling than that introduced by the NMMB/BSC-Dust model in the simulations. The methodology described for the dust model evaluation against the continuous P-MPL observations can be easily adopted for an operational use of the NMMB/BSC-Dust model for forecasting the mass concentration profiling in frequently dust-affected regions with serious climate and environmental implications, as long as a typical MEE for dust could be accurately specified. Hence, a statistical analysis for determining AERONET-based MEE values over the Iberian Peninsula is on-going.