8. Discussion and conclusions
In this paper, dust devil (DD) dust transport was investigated using LES. Our results show that the DD number density, n, decreases with Ri2 for negative Ri. n is smaller in our study than in the studies of Ohno and Takemi (2010) and Raasch and Franke (2011). The reason is likely the higher horizontal resolution used in the latter studies. For example, Raasch and Franke (2011) identi- fied twice as many DDs when using a 1 m horizontal resolution in their simulations rather than a 2 m grid resolution. Additionally, we excluded short tracks and connected tracks if pressure minima were found in the intermediate locations, thereby reducing our n. The order of magnitude of pressure drop, turbulent wind speed, and duration of the detected DDs are in agreement with those observed in the field (e.g. Sinclair, 1969; Metzger et al., 2011; Lorenz et al., 2015). With a mean value of hdi ¼ 86 m, the diameters in our study are somewhat larger than that of observed DDs (Balme and Greeley, 2006; Lorenz, 2011), with one reason likely being the different definition used to determine DD diameter. Another reason is probably an underestimation of the number of small DDs in our study. Our results show that instantaneous shear stresses in DDs are sufficiently large to aerodynamically lift dust particles as has been observed by Balme et al. (2003). Surface dust fluxes of about 100 103 lg m2 s1 and 2 m PM20 dust concentrations of 100 103 lg m3 have been obtained in our study, smaller or on the lower end of those observed in laboratory experiments and in the field. Neakrase and Greeley (2010) measured total mass removals of 105 1011 lg m2 s1 for 2 lm clay in their laboratory vortex generator experiments, much larger than the fluxes obtained here. In the field, Metzger et al. (2011) measured PM10 concentrations of 103 104 lg m3 at 2 m height, close to those obtained in our study for the more intensive DDs.
