1. NMMB-DUST developments
Carlos Pérez
29 July 2008
Includes preprocessing and first succesful model tests with:
- Dust emission with viscous sublayer
Dust horizontal advection
Dust vertical advection
Horizontal diffusion
Vertical diffusion
Also some notes, comments and future steps

2. Dust preprocessing in the NMMB-DUST

3. Dust calculations in preprocessing
Surface concentration (instead of emission) is the basis of viscous sublayer emission scheme during model integration
Surface concentration
Tunning parameter
Friction velocity
Threshold Friction velocity
Dry Threshold Friction velocity
Soil moisture
Soil water potential
Parameter function of
particle Reynolds number
Particle radius
Particle density
(1-Veg)
air density
Landuse fraction (desert mask)
Mass available for uptake (fraction
~source size distribution)
Calculated in preprocessing
fractions of clay (bins 1 to 4)
or silt (bins 5 to 8)
Calculated during model integration
Veg
Vegetation fraction (NEW) k
Size bin number (total is 8)

4. MAPS produced in preprocessing
Interpolation from initial database format to NMM B grid
In this example NMM B grid global at ~50 km resolution

5. Land use dust producing categories
2 LOW SPARSE GRASSLAND
8 BARE DESERT
11 SEMIDESERT
50 SAND DESERT
51 SEMIDESERT SHRUBS
52 SEMIDESERT SAGE

6. Land use fraction (0-1) Set to 1 at this point (future tunning)
if(landusedust(i,j).eq.2) landusefrac(i,j)=1.0 !LOW SPARSE GRASSLAND (DREAM 0.2)
if(landusedust(i,j).eq.8) landusefrac(i,j)=1.0 !BARE DESERT
if(landusedust(i,j).eq.11)landusefrac(i,j)=1.0 !SEMIDESERT (DREAM 0.8)
if(landusedust(i,j).eq.50)landusefrac(i,j)=1.0 !SAND DESERT
if(landusedust(i,j).eq.51)landusefrac(i,j)=1.0 !SEMIDESERT SHRUBS (DREAM 0.7)
if(landusedust(i,j).eq.52)landusefrac(i,j)=1.0 !SEMIDESERT SAGE (DREAM 0.7)

7. Vegetation fraction (0-1) NEW
Vegetation fraction (0-1) NEW!! Daily interpolated from monthly climatology.
Not existing in DREAM

8. 1-Vegetation fraction (0-1)

9. Top soil type (New STASGO database 30s… previous DREAM
Class No.
Soil texture class 1
Sand 2
Loamy Sand 3
Sandy Loam 4
Silt Loam 5
Silt 6
Loam 7
Sandy Clay Loam 8
Silty Clay Loam 9
Clay Loam 10
Sandy Clay 11
Silty Clay 12
Clay
Top soil type (New STASGO database 30s… previous DREAM
was a combination of FAO (2min) and ZOBLER (1º) )

10. Soil Texture triangle to determine the fractions of clay and silt
Class No.
Soil texture class
% Sand
% Silt
% Clay 1
Sand 92 5 3 2
Loamy Sand 82 12 6
Sandy Loam 58 32 10 4
Silt Loam 17 70 13
Silt 85
Loam 43 39 18 7
Sandy Clay Loam 15 27 8
Silty Clay Loam 56 34 9
Clay Loam
Sandy Clay 52 42 11
Silty Clay 47
Clay 22 20

11. Clay Fraction (0-1) Class No. % Clay 1 3 2 6 10 4 13 5 18 7 27 8 34 942 11 47 12 58
Clay Fraction (0-1)

12. Silt Fraction (0-1) Class No. % Silt 1 5 2 12 3 32 4 70 85 6 39 7 15 856 9 34 10 11 47 20
Silt Fraction (0-1)

13. w´ Soil water potential (volume fraction (%))

14. DELTA * Tunning parameter ( * )
Shown for k=4
(kg * s2 /m5)

15. 1 2 3 4 5 6 7 8 Dry threshold friction velocity (m/s)
Comes from old Bagnold’s theory which is valid for coarse particles (>30-40um radius)
But in fact, the friction velocity for saltation of fine particles should increase due to cohesive forces
Theory says that fine particles are uplifted by sandblasting (not saltation) which is not taken into account by DREAM parameterization with viscous sublayer. Instead, in DREAM the decreasing function of Balgnold
uplifts the fine particles by saltation (which is strictly incorrect if we believe in Saltation/Sandblasting Theory). We need to conciliate viscous sublayer approach with saltation/sanblastingtheory. If not possible we will have to switch to a flux form emission scheme taking into account saltation and sandblasting.

16. BSC NMMB-DUST (V200712 with old ESMF)
First successful test simulation including:
- Dust emission with viscous sublayer (PHYSICS)
- Dust vertical diffusion (PHYSICS)
- Dust horizontal diffusion (DYNAMICS)
- Dust horizontal advection (DYNAMICS)
Dust vertical advection (DYNAMICS)

17. In this version, dust is included in the existing components of the ESMF/NMMB
Experience has shown that switching to a new NMMB version would take 3-4 days.
This is the best way to develop at this moment.
Final version may include a separate module for dust emission, vertical diffusion
and deposition although this is not strictly necessary: it can be included as a subroutine
in the already existing physics component. Dust dynamics will be kept in the NMMB
dynamics component.
ATM
Coupler DYN-PHYS
Dynamics export
T, U, V, Q, CW, Q2, OMGALF, DUST
RUN DYN
(Now includes dust
hadv,vadv and hdiff)
RUN PHYS
(Now includes dust
emission and vdiff)
Coupler DYN-PHYS
Physics Export
T, U, V, Q, CW, Q2, DUST

18. PRELIMINARY ENCOURAGING RESULTS
Simulation set-up:
Global
Cycle UTC
60 hours forecast
160 CPU’s
8 dust bins
769x sigma layers

19. Cycle _12
54h forecast
Dust conc bin 4
Sigma layer 52
Kg/m3
Sorry for such a rudimentary visualization

20. Cycle _12
60h forecast
Dust conc bin 4
Sigma layer 52
Kg/m3

21. Next steps (I):
Finish a crude version of the model in the NMMB V version.
This means: include dry deposition (sedimentation and turbulent mix out)
and wet deposition as implemented in DREAM
Turbulent mix out parameterization from Giorgi 1986 in DREAM is a good option.
However I have checked the sedimentation scheme and compared
to other models (Chimere, Gocart) and needs to be changed (we need to talk about it).
We should remember also that wet deposition is very very simple and should be changed
as well in future versions
Once everything works in V200712: switch to the new V (includes new ESMF
Version). Also includes new tracer advection schemes (Old ones are lagrangian. New
ones are Eulerian, positive definite, mass conservative and monotonic). This change will
be straightforward.
Once a full version of the system works, Karsten can go on with his thesis and
improve/change emission, deposition schemes.

22. Next steps (I):
Other things to be done in the near future (list is not exhaustive):
- tests in regional mode
- postprocessing of DUST (currently does not write through the ESMF write component),
visualization of the NMMB in general
scalability tests
improvement of parallel behaviour with introduced dust component ….