SECTION 8: Using Models to Forecast Dust Storms
Hi Carlos. Your section is a good starting place but I think we need to make the information less technical and
more operationally oriented for weather forecasters. I think the last section of the AFWA Forecasting Dust Storms
module does a nice job of outlining the issues involved with dust models. Can you review it and see what you
think? (I’ve pasted the section below.) It’d be good to start with it, adding any important information that you think
is missing. We’ll need to substitute the case examples since we’re not including the case in our module. We
should definitely use your lists of dust models, adding short descriptions and links as you suggested. (Let me
know if it’s really time-consuming to assemble that information.)
Can you clarify the following for me? Are there two broad types of dust-related models?
1. Specific dust models -- that either produce dust forecasts or that forecast parameters related to dust. I
assume these are used for input into the more general models. Do weather forecasters ever see or use
these data/forecasts?
2. General NWP models with a dust component. Forecasters use the fields from these models to forecast
for areas that are expecting or experiencing dust events. Are forecasters aware of the dust versions of
these models? Do we need to teach them about them?
If there are 2 broad types of models, we should describe them at the beginning, noting what each is typically used
for (and if/when forecasters use them).
Proposed outline:
Section 1: Dust Forecasting With Models
Page 1: Different Types of Dust Models
Page 2: Identifying Dust Sources
Page 3: Model Dynamics & Dust Removal Processes
Plus any pages that you add
Page 4: Dust Models
Section 2: Dust Forecast Process
Page 1: Overview
Page 2: Long-range Forecast Process
Page 3: Medium-Range Forecast Process
Page 4: Short-Range Forecast Process
Page 5: Applying the Process to a Case
Here’s the text/graphics from the AFWA module’s model section (which I made some changes to). Note
that it comes after the case study, so it refers to things mentioned earlier.
Page 1: Different Types of Dust Models
The process of forecasting dust storms and surface visibility depends largely on model forecasts, which can differ
widely. In this section, we'll examine the main factors that account for these differences. These include the
models' processes for identifying dust sources, their dust transport dynamics, and their dust removal processes.
Following this, we’ll describe the major models used to forecast dust and then present a process for forecasting
dust storms.
Carlos: This would be the place to add the info on the 2 broad types of models.
1
Page 2: Identifying Dust Sources
The most critical factors in differentiating dust model forecasts are how dust sources are identified and the
resolution of the data. Some models get their dust source information only from satellites…
…while others use a combination of satellite, topographic, and land surface data, station data, atlases, and soil
samples.
For example: [I substituted the DOD model examples with yours]

The NAAPS model bases its dust sources on the observed frequency of high aerosol values of [OR:
from?] the Total Ozone Mapping Spectrometer (TOMS) Aerosol Index.
2

The NMMB/BSC-Dust model from the Barcelona Supercomputing Center uses a function based on
topography and a land cover map to identify major dust sources

Finally, the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast
System uses the ultraviolet-visible component of land surface albedo [or can we just say: ultravioletvisible albedos?] to diagnose areas likely to produce a flux [clarify word] of dust.
The important thing is how a model determines the number and extent of dust sources in each grid box. If a
model 'thinks' that many dust sources cover a significant portion of a grid box, it may predict large, broad plumes
for the area. Another model may not show any dust sources for the box. If that's not correct, it may reflect a
weakness in how the model determines erodible or dust-producing areas.
3
[if we need examples, we should change the ones shown here or modify the discussion so it doesn’t mention the
case] The models used in the 26 February case have varied dust source functions. Some forecast broad plumes,
others narrow plumes. Some have too many dust sources, others too few.
For example, COAMPS forecasted several refined plumes for the interior of the UAE and its coast...
… while DTA-WRF forecasted a broad dust plume with low visibilities (1 to 0.5 miles or 1.6 to 0.8 km) from Qatar
to the Strait of Hormuz. The COAMPS forecast was based on a few, limited dust sources, whereas DTA-WRF
had too many.
4
Over the Saudi Plateau, COAMPS over-predicted surface dust concentrations, leading to a broad area with
visibilities from 2 to 0.5 miles (3.2 to 0.8 km). In contrast, DTA-WRF did not forecast any reduced visibility plumes.
This suggests that COAMPS had too many dust sources for this area, while DTA-WRF had too few.
Carlos: Here’s the info from your section. Add anything you think needs to be said.
Land use type and the location of sources
Traditionally, models have identified [had said located] dust sources based on the bare ground categories
in land cover maps. For example, the NAAPS global dust model used to identify sources regions based
on the USGS global database of land cover types. As described in the dust source section, the database
has eight dust-producing land covers: low sparse grassland in Mongolia; bare desert south of 60°N; sand
desert; semi-desert shrubs south of 60°N; semi-desert sage; polar and alpine desert; salt playas/sabkhas;
and sparse dunes and ridges. [If this is no longer right, we need to change it in the dust source
section!!!!!!!!!!!!]
[Carlos: The following information is used in the section of the module on dust source regions. Is it OK to
keep it there or is it considered outdated?] Recent analyses of satellite data have shown that most major
sources are located in topographic depressions where deep alluvial deposits have formed.
Current model representations of major sources are based on topographic, hydrologic, and/or
geomorphologic factors. Or they are derived directly from satellites using surface reflectance, frequency
of high aerosol values, and ultraviolet-visible albedos. This produces a far more refined view of global
dust source regions.
ORIG: More recently, satellite data analysis has shown that most major sources are located in
topographic depressions where deep alluvial deposits have formed. There are different model
representations of major sources based on topographic, hydrologic or geomorphologic
considerations, or directly derived form satellites like surface reflectance, frequency of high
aerosol values, and UV-visible albedos. This results in a much more refined view of global dust
source regions. Several examples are:
[this was followed by your examples. I’ve added the text but not the graphics since they’re shown above]
For example, the NAAPS model bases its dust sources on the observed frequency of high
aerosol values of [from?] the Total Ozone Mapping Spectrometer (TOMS) Aerosol Index.
Other models, such as the NMMB/BSC-Dust at the Barcelona Supercomputing Cente, use a
function based on topography and a land cover map to identify the major dust sources.
The European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast
System uses the ultraviolet-visible component of land surface albedo [OR CAN WE JUST SAY:
ultraviolet-visible albedos??] to diagnose areas likely to produce a flux [clarify that word] of dust.
Model Dynamics & Dust Removal Processes
NWP models produce different dynamical forecasts due to their sensitivity to initial conditions. This can lead to
different atmospheric motions and stability, which can create variations in the strength and location of upper-level
5
short waves, surface lows, associated fronts, and surface winds. Differences in the forecasted strength and
location of surface winds account for different dust visibility forecasts among models.
Finally, a model's handling of soil moisture and precipitation impacts its treatment of dust production and removal.
Source areas with significant rainfall in previous time steps will have high soil moisture values and suppress dust
production in current forecasts. Since rain removes suspended dust particles, differences in forecasted
precipitation patterns lead to different visibility forecasts.
Page 4: Dust Models
Carlos: Yes, for each model, can you add a brief description, an example (if possible), and a link to the site?
Introduce this part on dust models.
Here’s info from your section below. Decide if it fits here.
The use of dust model forecasts depends on the time range of the forecast. Short-range dust forecasts tend to
rely on real-time analyses, while medium- and long-range forecasts rely far more on model output from
mesoscale and global-scale models.
Mesoscale models provide high-resolution regional dust predictions generally up to three days while global-scale
models provide lower-resolution global dust predictions up to five or six days.
FIND A HOME FOR THIS!! Models generally do a good job of reproducing the spatial and temporal variability of
dust when dust emission is caused by large-scale dynamics, such as pre- and post-frontal winds. However, they
usually misrepresent dust emissions related to moist convective events, such as haboobs. The use of wind fields
from mesoscale models can improve forecasts for certain scenarios that require high-resolution data, such as
dust storms promoted by gap flows. [ORIG SENTENCE: The use of wind fields derived from mesoscale models
can improve forecasts for certain scenarios that require high resolution such as dust storms promoted by gap
flows.
The following dust models provide real-time information about ???.
Mesoscale models:
Model
Source
COAMPS, DTA-MM5,
and DTA-WRF
BSC-DREAM8b and
NMMB/BSC-Dust
(regional)
Chimere-DUST
U.S. Department of
Defense
Barcelona
Supercomputing
Center (BSC)
IPSL/Laboratoire de
Météorologie
Dynamique (LMD)
University of Athens
SKIRON
CUACE/DUST
MASINGAR
ADAM
Description [need
this?]
Area of coverage
Link
Northern Africa,
Middle East, Europe
Northern Africa,
Middle East, Europe
Northern Africa,
Middle East, Europe
East Asia
China Meteorological
Administration (CMA)
Japan Meteorological
Agency (JMA)
Korea Meteorological
Administration (KMA)
East Asia
East Asia
Global-scale models:
DTA-GFS, NAAPS,
NOGAPS
IFS-GEMS
U.S. Department of
Defense
ECMWF
Description
Link
6
GFS-GOCART
NMMB/BSC-Dust (global)
MOCAGE
(European Center
for Medium range
Weather
Forecasting)
NCEP (National
Centers for
Environmental
Prediction)
BSC
MeteoFrance
CARLOS: Here’s your section, with some changes made. Incorporate what you think is useful!
8.1 Dust modeling
8.1.1. Dust models
8.1.2. Dust source and emission modeling
8.1.3 Types of models (i.e. global and regional models)
8.1.3 Strengths and limitations of dust models
8.1.4 Available dust models
8.2 Dust forecasting
8.2.1 Introduction
8.2.2 General considerations when making a dust storm forecast
8.2.3 Long range
8.2.4 Medium range
8.2.5 Short range
8. Dust Forecasting With Models
This section provides an overview of dust forecast models, describing the different types of models [OR:
describing some of the major models??], their period of coverage [clarify!] and their strengths and limitations. This
sets the stage for the final section, where we present a process of forecasting dust storms. As you know, model
forecasts are used extensively during the medium- and long-range forecast periods and far less so during the
short-term and nowcasting periods.
8.1 Dust modeling
8.1.1. Dust models
Dust models combine dust source information with predictions of the atmosphere to forecast the occurrence of
dust storms. Dust models essentially simulate the atmospheric dust cycle, from [arrange the following items in
order of the cycle!] dust emissions from source areas to … turbulent diffusion, advection, convection,
sedimentation, and dry and wet deposition.
The key component of a dust forecast model is how it treats dust sources and emissions. If a model cannot
adequately predict the production phase of the dust cycle, all of its other dust processes will be misrepresented.
8.1.2. Dust source and emission modeling
Currently, scientists do a good job of computing the vertical flux of dust in erodible, arid surfaces [areas?] if the
input parameters are accurate [had said: determined accurately]. However, emission schemes are often
hampered by inaccurate data or data that’s not available at the right scales. To address these problems, models
simplify the dust emission scheme to varying degrees based on the amount and accuracy of the available data.
Current theoretical knowledge allows computing satisfactorily the vertical dust flux in erodible arid
surfaces provided that the required input parameters are accurately determined. However, the application
of emission schemes is mainly hampered by the inaccuracies or the lack of required input data at the
7
pertinent scales. In order to partly underwhelm these limitations, current models assume varying degrees
of simplification in the dust emission scheme as a function of available data.
Dust emission schemes use a variety of input data, such as land use types, source locations , friction velocity, soil
texture, soil humidity, and the surface roughness length [define]. We’ll look at these in more detail.
Land use type and the location of sources
Traditionally, models have identified [had said located] dust sources based on the bare ground categories in land
cover maps. For example, the NAAPS global dust model used to identify sources regions based on the USGS
global database of land cover types. As described in the dust source section, the database has eight dustproducing land covers: low sparse grassland in Mongolia; bare desert south of 60°N; sand desert; semi-desert
shrubs south of 60°N; semi-desert sage; polar and alpine desert; salt playas/sabkhas; and sparse dunes and
ridges. [If this is no longer right, we need to change it in the dust source section!!!!!!!!!!!!]
Recent analyses of satellite data have shown that most major sources are located in topographic depressions
where deep alluvial deposits have formed. [Carlos: This information is used in the section of the module on dust
source regions. Is it OK to include it or is it considered outdated?]
Current model representations of major sources are based on topographic, hydrologic, and/or geomorphologic
factors. Or they are derived directly from satellites using surface reflectance, frequency of high aerosol values,
and ultraviolet-visible albedos. This produces a far more refined view of global dust source regions.
ORIG: More recently, satellite data analysis has shown that most major sources are located in
topographic depressions where deep alluvial deposits have formed. There are different model
representations of major sources based on topographic, hydrologic or geomorphologic considerations, or
directly derived form satellites like surface reflectance, frequency of high aerosol values, and UV-visible
albedos. This results in a much more refined view of global dust source regions. Several examples are:
[i added the following examples above]
For example, the NAAPS model bases its dust sources on the observed frequency of high aerosol values of
[from?] the Total Ozone Mapping Spectrometer (TOMS) Aerosol Index.
8
The NMMB/BSC-Dust model from the Barcelona Supercomputing Center uses a function based on topography
and a land cover map to identify the major dust sources.
The European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast System uses the
ultraviolet-visible component of land surface albedo [OR CAN WE JUST SAY: ultraviolet-visible albedos??] to
diagnose areas likely to produce a flux [clarify that word] of dust.
Friction velocity
As you’ll recall, dust storm generation requires high wind, wind shear, and turbulence, and unstable boundary
layers. The friction velocity parameter derived from NWP models expresses wind speed, turbulence, and stability
and is the key meteorological parameter [had said variable] in dust emission modelling.
Because dust emission is a threshold phenomenon, models use threshold friction velocities above which dust
emission can occur. These may assumed to be constant or depend on soil features such as texture, humidity, and
roughness. [ORIGINAL: aeolian roughness lengths. ALTERNATIVE: roughness due to wind-generated geologic
processes]
According to the threshold wind friction velocity of soil particles, the optimum particle size for uplift ranges from 60
to 80 micrometres. The threshold friction velocity is greater for smaller particles such as clays and fine silts due to
the cohesion forces between particles. It’s also greater for larger particles, such as sand, due to gravity. [simplify
those statements!]
CHECK AGAINST THE ORIGINAL: Because dust emission is a threshold phenomenon, models use
threshold friction velocities above which dust emission can occur. Thresholds may be assumed as
constant or depending on other soil features, for example, soil texture, soil humidity and aeolian
roughness lengths.
9
The threshold wind friction velocity of soil particles shows an optimum particle size range for uplifting
due to inter-particle cohesion forces and for larger particles (sand) the threshold friction velocity increases
due to gravity. Most of the dust is suspended through sandblasting or saltation bombardment, the process
by which saltating grains break the inter-particle bonds linking dust particles together or to the surface.
Where does this go??? Most dust is suspended through sandblasting [clarify] or saltation bombardment in which
the wind breaks the bonds that link grains to each other or to the surface [ok?].
Soil texture
Some simple models assume a constant soil size , while many others calculate the emission flux [simplify
emission flux] using estimates of soil size distribution. These maps show estimates of mass fractions of clay, silt,
fine-medium sand, and coarse sand derived from the STATSGO-FAO soil texture database. briefly explain the
impact of using these.
Soil humidity
Soil water can inhibit dust emission by increasing the cohesion between soil particles. Dust?? models typically
inhibit dust emission depending on the soil moisture provided by the NWP model. Note that soil moisture largely
depends on precipitation, which NWP models still have a difficult time predicting.
10
Surface roughness length
Some models have begun using satellite-derived estimations of roughness length to account for the fact that nonerodible elements, such as pebbles, stones, and vegetation, dissipate some wind momentum. This reduces the
wind available to promote saltation and dust emission.
8.1.3 Types of Models and Their Limitations (DISCUSS ABOUT POTENTIAL FIGURES)
[I added this to the section above] The use of dust model forecasts depends on the time range of the forecast.
Short-range dust forecasts tend to rely on real-time analyses, while medium- and long-range forecasts rely far
more on model output from mesoscale and global-scale models.
[I added this to the section above] Mesoscale models provide high-resolution regional dust predictions generally
up to three days while global-scale models provide lower-resolution global dust predictions up to five or six days.
Due to the relationship between wind and dust, small errors in surface friction velocity from NWP models leads to
large errors in dust emissions. This makes the accuracy of the wind speed used in dust models critical. Surface
wind speeds in models tend to be underestimated over source regions and not account for specific meteorological
situations that are responsible for dust emissions, such as density currents and squall lines.
But models generally do a good job of reproducing the spatial and temporal variability of dust when dust emission
is caused by large-scale dynamics, such as pre- and post-frontal winds. However, they usually misrepresent dust
emissions related to moist convective events, such as haboobs. . The use of wind fields from mesoscale models
can improve forecasts for certain scenarios that require high-resolution data, such as dust storms promoted by
gap flows. [ORIG SENTENCE: The use of wind fields derived from mesoscale models can improve forecasts for
certain scenarios that require high resolution such as dust storms promoted by gap flows.
8.1.4 Dust Models (put in section above]
Over the last two decades, several dust models have been developed that provide real-time information about
???.
The mesoscale models include:







COAMPS, DTA-MM5, and DTA-WRF from the U.S. Department of Defense (ADD REGION)
BSC-DREAM8b and NMMB/BSC-Dust (regional) from the Barcelona Supercomputing Center (BSC) for
Northern Africa, the Middle East, and Europe
Chimere-DUST of the IPSL/Laboratoire de Météorologie Dynamique (LMD) ) for Northern Africa, the Middle
East, and Europe
SKIRON from the University of Athens for Northern Africa, the Middle East, and Europe
CUACE/DUST from the China Meteorological Administration (CMA) for East Asia
MASINGAR from the Japan Meteorological Agency (JMA) for East Asia
ADAM from the Korea Meteorological Administration (KMA) East Asia
The global-scale models include:





DTA-GFS, NAAPS, NOGAPS from the U.S. Department of Defense
IFS-GEMS from ECMWF (European Center for Medium range Weather Forecasting)
GFS-GOCART from NCEP (National Centers for Environmental Prediction)
NMMB/BSC-Dust (global) of the BSC
MOCAGE of MeteoFrance
THIS WILL BE OUR NEW SECTION 2
Section 2: Dust Forecast Process
Page 1: Overview
11
This section presents a general process for forecasting dust storms that incorporates the wide array of tools
available to help forecasters predict dust storms. These tools include satellite imagery and RGB products, surface
and upper-air observations, NWP models, and a new generation of dust/aerosol models. [are all of them referred
to in the process?]
The dust forecast process is divided into three parts defined by the forecast lead time:



Long range, 72 to 180 hours
Medium range, 24 to 72 hours
Short range, 0 to 24 hours
Before starting to develop a dust forecast, forecasters should be familiar with their area of responsibility and local
rules of thumb. In particular, they should know:






The types and locations of local dust source regions; for example, if there are lake beds, salt flats, or newly
developed drought regions
The types of soil present
The impact of local terrain on wind speeds
The wind direction with respect to local dust source regions
How the winds align from the upper levels down to the surface (vertical wind shear), especially during winter
If available the dust climatology of the region of interest; for example from models, satellites and/or other dust
observations
If you are a forecaster, here are some tips to keep in mind when viewing model guidance. When possible:


Consult different dust products from the same dust model or a different model since each product provides
slightly different information
Animate forecast products to identify mesoscale dust features and their movement, extent, and location
Page 2: Long-range Forecast Process
The long-range (72- to 180-hr or 3- to 5-day) dust storm forecast process has two steps.
1. Look for large-scale, synoptically driven dust events in the 3 to 7.5-day range in global models, such as, IFSGEMS, GFS-GOCART, NMMB/BSC-Dust, MOCAGE, DTA-GFS, and NAAPS
2. Look for model-forecast midlatitude troughs that drive pre-and post-frontal dust storms in winter and that can
amplify the large-scale wind patterns associated with summer events. These large-scale waves are resolved
by global NWP models such as GFS, IFS, NMMB, and NOGAPS. The associated dust outbreaks are
modeled by global dust models such as DTA-GFS, NAAPS, IFS-GEMS, GFS-GOCART, NMMB/BSC-Dust,
and MOCAGE
Page 3: Medium-Range Forecast Process
For the 24- to 72-hr forecast, use mesoscale dust model output from models such as the:
o BSC-DREAM8b
o NMMB/BSC-Dust
o Chimere-DUST
o SKIRON for Northern Africa, the Middle East, and Europe
o CUACE/DUST, MASINGAR and ADAM for East Asia
In addition, use larger-scale dust forecasts from models such as:
o Global DTA-GFS
o NAAPS
o IFS-GEMS
o GFS-GOCART
o NMMB/BSC-Dust
o MOCAGE
12
Guidance from NOGAPS, GFS, and IFS shows the evolution of larger-scale atmospheric features and is helpful
for identifying conditions favorable for a blowing dust event.
Here are the steps in the medium-range dust storm forecast process.
Step 1: Examine the following charts:



300-mb height and wind forecast charts to track troughs and jet streaks; briefly examine upper-tropospheric
winds to identify the presence of any jet streaks, especially for cool-season dust storms; jet streaks within a
pronounced upper-level trough are indicative of an intensifying low-pressure system with stronger surface
fronts and associated winds
500-mb height and relative vorticity forecasts to identify and track troughs and vorticity maxima
MSLP and surface wind forecast charts for fronts and potentially strong wind conditions.
Step 2: Looking at the forecast soundings from a model, determine the forecast stability and wind profile at your
forecast time of interest.
Step 3: Check the 6-hrly precipitation and 700-mb relative humidity forecast charts to determine where increased
moisture and precipitation are anticipated since they decrease the probability of dust lofting.
Step 4: Combine forecasts of surface friction velocity, surface winds, and soil wetness with your knowledge of
dust source areas to see if the criteria for a potential blowing dust event are met. Recall that friction velocity
incorporates atmospheric stability and wind speed into one variable.
Step 5: Examine forecasts of surface visibility and or surface concentration due to dust. Compare them to
forecasts of winds through the mixed layer and dust optical depth to help assess changes in geographical extent
and intensity with each successive model run.
Step 6: From the model output and your initial analysis, develop a best-guess forecast as to the onset and
duration of any dust events in your area of responsibility in the 24- to 72-hr window.
Page 4: Short-Range Forecast Process
The process for creating short-range (0- to 24-hr) dust forecasts includes the following steps.
Step 1: Analyze the present state of the atmosphere by looking at satellite imagery, upper-air charts, and surface
analyses, keeping in mind the location and characteristics of relevant dust source regions.
Step 2: Examine the latest observed and/or forecast soundings. Note the strength of any inversions (usually
during summertime) and determine if they will break due to turbulent mixing and daytime heating that would ripen
the environment for a dust outbreak.


A dry adiabatic lapse rate from the surface through a deep mixed layer allows the dust to loft to great heights,
especially if winds are from the same direction and increase with height through the layer
o
Note that dust storms generally occur in this kind of environment and that the strongest wind speed
aloft within the dry adiabatic layer can be brought to the surface
o
The height or top of an elevated dust layer can be approximated by determining where the lapse rate
becomes less than the dry adiabatic lapse rate
Dust storms are less likely in a stably stratified boundary layer although narrow plumes of blowing dust are
still possible
Step 3: To determine the potential duration and type of dust event, pay special attention to dust lofting in your
area of responsibility, local rules-of-thumb about advection, and geographic features such as the location of dust
13
source regions, terrain, vegetation, and water sources. Also note where precipitation has fallen in the past 48
hours and whether it was convective or stratiform.
Step 4: Use satellite dust enhancement products (such as enhanced infrared imagery) and RGB and other
multispectral imagery tuned for dust detection. Integrating these products with surface observations can provide
information about the current extent and location of existing dust plumes and fronts.
Step 5: Make a best-guess forecast as to the onset, duration, and persistence of any dust events in your area of
responsibility in the very short term, using short-range model from mesoscale models as guidance. The global
models can resolve large-scale features that drive smaller-scale dust events in the short term but cannot resolve
more localized dust features.
Page 5: Applying the Process to a Case
The steps that we’ve outlined are applied to a case in another COMET module, Forecasting Dust Storms. It, too,
contains general information about dust. However, the examples are mostly from the Middle East and the data
are from U.S. Department of Defense models. But since the case addresses a common type of dust storm, a
frontal event, it’s useful to go through it from a procedural and meteorological standpoint. To access the case, go
to http://www.meted.ucar.edu/mesoprim/dust/ and select the near-final section, Forecasting Dust Storms.
14