1. The Importance of the Representation of Deep Convection for Modelled Dust-Generating Winds over West Africa during Summer J. H. Marsham 1,2, P. Knippertz 2, N. Dixon 2, D. J. Parker 2 and G. M. S. Lister 1,3 1 National Centre for Atmospheric Science, UK. 2 University of Leeds, UK. 3 University of Reading, UK. Contact: j.marsham@see.leeds.ac.uk 10-day Unified Model (UM) simulations suggest that cold-pool outflows from convective storms (“haboobs”) generate ~50% of the dust-generating winds in summertime West Africa. Haboobs are poorly captured in models with parameterised deep convection.  To isolate the role of meteorology we define “uplift potential” as U 3 (1+U t /U)(1-U t 2 /U 2 ), where U is 10-m wind and U t =7 ms -1 a typical threshold velocity 2. Uplift potential would control parameterised dust uplift 1 if the bare-soil land surface was uniform.  Simulations with explicit deep convection show haboobs generating ~50% of the uplift potential (Fig 3), but 10-day averages of uplift potential are similar between all simulations (Fig. 2).  Over 10 days, the lack of haboobs in models with parameterised convection is compensated by increased uplift from nocturnal low-level jets (LLJ, Fig. 3b) resulting from of a stronger Saharan Heat Low (SHL) in these models (Fig. 4).  Only runs with explicit deep convection show significant uplift potential over the Sahel (Fig. 2), which may help explain an observed northwards bias in dust in some regional models 3.  Tuning cannot resolve the problem of haboobs in models with parameterised deep convection – a haboob parameterisation is required. References: 1 Marticorena and Bergametti, JGR,100, 1995. 2 Chomette et al., JGR, 104, 1999. 3 Johnson et al., early view QJRMS, 2011. 4 Pearson et al., JGR, 115, 2010. 5 Knippertz, Met. Zeitschr., 17, 2008. 6 Weisman et al., MWR, 125, 1997. 7 Bou Karam et al., QJRMS, 134, 2008. 8 Marsham et al., JGR, 113, 2008. Paper on this has now been published in GRL!!! The Cascade project simulations Four UM 4 runs for the 10-days 25 July to 03 August 2006: (i) 40-km and 12-km runs with parameterised deep convection (ii) 12-km and 4-km runs with explicit deep convection. Also, a 1.5-km explicit run for the 25 to 26 July. All initialised and forced at boundaries with ECMWF analyses. Fig. 1. UM domains.  Main maximum at southern edge of SHL (16 to 22°N) in all runs.  Area totals in [ ] similar in all runs.  Arcs from haboobs in runs with explicit deep convection (most apparent in the Sahel). 12 km Explicit [15.0x10 5 ] 4 km Explicit [15.4x10 5 ] 12 km Param. [15.1x10 5 ] 40 km Param. [16.3x10 5 ] E.g. Arcs from haboobs  09–10 UTC mixing of low-level jet (LLJ) to the surface 5 in all runs.  Haboob peak in the afternoon accounts for ~50% in explicit runs, is greatly underestimated with parameterised convection and is delayed for coarser grid spacings as expected 6.  10-day runs: lack of haboobs with parameterised deep convection compensated by stronger SHL (Fig. 4; more LLJs and increased nocturnal monsoon flow 7 ).  Simply using 6-hourly wind analyses would miss both LLJs 5 and haboobs 8. Haboob peak LLJ peak Haboob peak Dashed=param. Solid=explicit Dashed=param. Solid=explicit Nocturnal monsoon flow 1.5-km domain Fig. 4. The mean SHL  Runs with parameterised deep convection develop a deeper SHL and so a stronger LLJ (Fig. 3b),  All simulations are within two analyses.  More observations from SHL region needed.  Explicit cold pools may lead to more effective ventilation of SHL. 40 km param. 12 km param. 12 km explicit 4 km explicit ECMWF analysis NCEP analysis Fig. 2. Ten-day mean uplift potentials Fig. 3. Mean diurnal cycle Mean: 25-26 July Mean: 25 July to 3 Aug. 4-km domain