Extremely low PM10 by CMAQv5.3

Hi,
I wonder anyone has experience to conduct PM10 simulation using CMAQv5.3.
I got extremely low PM10 concentrations in entire model domain by CMAQ5.3 from September 1 to 30, 2017. Emissions include precursors of secondary organic aerosols (SOA) and is based on EPA 2017 NEI datasets. I turned on windblown dust options in the run script, and set BELD3 for DUST_LU_1 and DUST_LU_2. WRFv4.2/MCIPv5.0 outputs show some high wind days are available in this month. The chemistry mechanism is cb6r3_ae7_aq. The results from CMAQ post-processor combine show that PM10 is extremely low comparing against observational data within entire domain. Please see attached for the result at one monitor site. However, modeled ozone concentrations (hourly and 8-hourly) are in good agreement with observations.

Some key settings in the run script :
set MECH = cb6r3_ae7_aq
setenv CTM_WB_DUST Y
setenv CTM_WBDUST_BELD BELD3
setenv CTM_DUSTEM_DIAG Y

Hello,
I am wondering if anyone has experience to estimate windblown dust or PM-10 using CMAQ v5.3 and CMAQ v5.3.1. I would appreciate if you share your experience such as input preparation and results.
Thanks.