Particulates, PM10, PM2.5 and increasingly PM1.0 become pollutants of primary concern in many cities and regions world wide. Health effect of fine particles that enter the respiratory pathways are reason for concern, effective control measure hotly disputed in particular where they involve restriction on individual passenger cars, access to inner city areas.

In many areas, however, dust from natural sources (wind entrainment of soils) may play a dominant role, together with an important contribution from long-range transport. EPA estimates that more than 80% of dust emissions (over the continental US) originate from miscellaneous natural sources (21%) and fugituve dust (67%), with only 12 % directly attributable to human activities, fuel combustion, other industrial processes, and behicles.

One of the important questions as the basis for any effective control measures is a reliable source apportionment between antropogenic, largely but not exclusively combustion related (and thus in principle controllable sources), and natural sources, that is wind erosion of soils, entrainment and re-suspension including very long range transport, where efficient control options are more elusive.

AirWare include a dynamic emisison model for wind erosion of soils, but also from activities such as demolition and construction activities, unpaved roads, surface mining, and many agricultural activities such as plowing and tilling or the use of harvest combines.

Operational forecasts, nowcasting
The particulate modeling is based on daily model/forecast runs over a 3-7 day forecast horizon, including:

• Downloading and pre-processing of NCEP/GFS global forecast data for the a large region around the primary area of interest, 5 to 10 times the horizontal extent in multiple layers of "nesting".
• Dynamic downscaling of the forecast (3 level nesting) with MM5;
• Running wind erosion model for the region (based on a two parameter Weibull function logarithmic wind speed distribution around MM5 hourly means, and a logarithmic threshold function of wind speed and (dynamic) erodibility based on vegetation cover, land surface, and soil/moisture data, see below;
• PM10/2.5 transport simulation with 3D Eulerian model CAMx, nested grid (two levels of nesting, two-way coupling, master and national/local domain); Model output (available for download from a dedicated web server)
• Daily updates of 3-7 day forecasts (hourly) of PM10/2.5 concentration and deposition data, regular model grid and derived statistical and graphical interpretation for up to 10 vertical layers.
• Continuous system logs and regular performance reports, self-check with real-time expert system RTXPS for on-line QA/QC, error detection, operator alerts (mail, SMS).

Historical (re-analysis) model runs:
To calibrate the model system, first step for any regional application is to re-run previous years with the latest re-analysis data to determine (historical) source areas and source contributions (possibly with historical landuse data derived from LANDSAT RS imagery). There are however several different versions of reanalysis (NCEP1, NCEP2, ERA40, ERA40 interim). Later versions usually have improved method or new data included. Reanalysis data from NCEP/NCAR are available for 1870 up to 2008: http://dss.ucar.edu/datasets/ds131.1/

The FNL data are still available from 1948 till present day. Probabilistic contribution analysis and dust mapping by super-ensembles use of several data sets and models (NCEP1, NCEP2, ERA40, ERA40interim, different models: MM5, WRF) data for probabilistic analysis. Alternative meteorological model configurations, using, for example:

• different pbl model, soil moisture model, numerical scheme, cloud
• parameterization, vortex following nesting (in WRF), etc. (also applicable for the real-time forecast runs).

Model input data compilation and processing
The cascade of the three simulations models (MM5, Emission model, CAMx) uses the following spatial inputs, compiled and pre-processed from 1 km resolution data sources to the nested model domain resolutions of 9, 3, and (optional) 1 km. The data are projected to the UTM zone of the center domain with progressively larger errors farther away from the domain center.

Basic land data for MM5 (topography, land use, soil, vegetation) can be taken from any national data base or global USGS datasets. More information about the default datasets can be found at http://eros.usgs.gov/#/Find_Data/Products_and_Data_Available/gtopo30/README

• 5 min ( ~9 km) global terrain and landuse for main domain (27 km)
• 2 min ( ~4 km) global terrain and landuse for 1st nested domain (9 km)
• 30 sec ( ~.9 km) global terrain and landuse for 2nd nested domain (3 km)

For dynamic forcings, AirWare can use and post-processe NCEP/NCAR products:

• GFS (forecast): http://wwwt.emc.ncep.noaa.gov/gmb/moorthi/gam.html
• GEFS (ensemble): http://products.weather.gov/PDD/NCEPMAF.pdf
• FNL (reanalysis): http://dss.ucar.edu/datasets/ds083.2/.

DUST emission model and background maps:

• DEM: 1 km or (optional) 30 m resolution data sources http://free-gis-data.blogspot.com/2009/04/aster-global-digital-elevation-model.html
• Landcover, satellite derived
  • Historical: http://edc2.usgs.gov/glcc/glcc_version1.php#Eurasia
  • Recent: http://glovis.usgs.gov
• Soil data): http://www.fao.org (1:5M vector map) or local soil maps/data

• Vegatation data:
  • NDVI (MODIS) https://lpdaac.usgs.gov/lpdaac/products/modis_overview with optional dynamic (seasonal, monthly) NDVI data, interpolated.
  • VCF, Vegetation Continuous Fields,
  The Vegetation index (NDVI), directly obtained from remote sensing data or land cover/ land use (e.g., CORINE or USGS classification (see: http://edc2.usgs.gov/glcc/afdoc2_0.php#olso (chapter 4.2) , with an NDVI range and default value associated with every land cover class); the NDVI estimates can be derived from MODIS satellite data (500 m resolution, 32 day average reflectance data). An alternative is a global data set from the Global Land Cover Facility (www.landcover.org) that covers the percentage of woody vegetation, herbaceous vegetation, and a bare ground percentage (source: http://glcf/umd.edu/data/vcf).

  The Vegetation Continuous Fields collection contains proportional estimates for vegetative cover types: woody vegetation, herbaceous vegetation, and bare ground. The product is derived from all seven bands of the MODerate- resolution Imaging Spectroradiometer (MODIS) sensor onboard NASA's Terra satellite. GLCF editions of MODIS products differ from DAAC editions by coming in GeoTIFF format, geographic coordinates, WGS84 datum, and a tiling system designed to fit well with Landsat imagery. The continuous classification scheme of the VCF product may depict areas of heterogeneous land cover better than traditional discrete classification schemes. While traditional classification schemes indicate where land cover types are concentrated, this VCF product is great for showing how much of a land cover such as "forest" or "grassland" exists anywhere on a land surface.

DUST: Wind entrainment model basics

The dust entrainment model is a distributed, (arbitrary resolution) dynamic (hourly time step) model to predict the wind erosion and entrainment of particles from natural surfaces, primarily unvegetated or sparsely vegetated soils.

The model produces dynamic (hourly) emission matrices, that together with any anthropogenic emission data for point, line and area sources, provide input to the respective transport, dispersion, and deposition model used (e.g., nested grid CAMx).

In addition to the threshold friction velocity approach (e.g., Draxler et al. 2001) that uses geomorphology and soil properties, is also considers vegetation cover (see above), soil moistur, soil typse, and slope and aspect of of model grid cell. Soil moisture is estimated by the MM5 prognostic meteorological model. MM5 is also used to forecast wind velocities, and a Weibull function to generate distributed wind speeds around the predicted hourly mean wind speed.

The dust entrainment model estimates non-pyrogenic dust emission from natural surfaces as a function of primarily wind speed, land cover/vegetation, soil characteristics, and soil moisture, The total Dust PM10 emission [g/s/ha or km2] is calculated from

• WindFactor (generated with a two-parameter Weibull function from hourly mean wind speed generated by MM5 (optional WRF), down to 1km diagnostic interpolation),
• ErosionFactor (erodibility, using a vegation index, soil type, soil moisture, slope and aspect)

Wind entrainment is a non-linear threshold function of wind speed and the erosion factor.