A Dynamic Adaptive Grid Method for Improved Modeling of Biomass Burning Plumes

1
A Dynamic Adaptive Grid Method for Improved
Modeling ofBiomass Burning Plumes

• M. Talat Odman and Yongtao Hu
• Georgia Institute of Technology
• D. Scott McRae
• North Carolina State University
• Gary L. Achtemeier
• USDA Forest Service
• 7th Annual CMAS Conference
• 6 October, 2008

2
Objective

• To improve the prediction of air quality impacts
  from biomass burnings.
• The focus is on prescribed burns and their
  impacts on local and regional air quality.

Simulation with Air Quality Model
Burning Options
Prescribed Burns at Managed Lands
Impact to Regional Air Quality
3
Approach

• Characterization of Emissions and Air Quality
  Modeling for Predicting the Impacts of Prescribed
  Burns at DoD Lands

4
Model Development

• Biomass burning plumes are not well resolved in
  current regional-scale modeling systems due to
• insufficient grid resolution
• inadequate subgrid treatments
• A dynamic, solution-adaptive grid algorithm
  (DSAGA) will be used both in MM5 and CMAQ models
  to increase their resolutions.
• Turbulence parameterization has already been
  revised in MM5. A subgrid scale plume
  model will be coupled with CMAQ.

5
Dynamic Adaptive Grid Resolve selected
features/characteristics/properties dynamically
6
Incorporating DSAGA into CMAQ Time Stepping

• Process modules are called once every ?t
• There is one global ?t for the entire domain
• Since ?x is constant max(u) determines ?t
• For non-uniform grids, min(?x) and max(u)
  determine ?t

7
VARTSTEP Algorithm

• Every cell is assigned its own local time step,
  ?ti ,
• which is an integer multiple of the global time
  step ?t and an integer divisor of 60 minutes. For
  example., if the global time step is 5 minutes,
  the local time step can be 5, 10, 15, 20, 30, or
  60 minutes.
• The model clock time, t, is advanced by the
  global time step
• When tN??ti processes are applied for the
  duration of ?ti
• Transport requires special attention.

8
Rotating Cone Test
9
January 1-9, 2002 Simulation
Dx 12 km
CPU savings with VARTSTEP 25
10
Adaptive Grid MM5

• NCSU Dynamic Solution Adaptive Grid Algorithm
  (DSAGA)
• (r-)Refinement criteria selected beforehand
  (currently vorticity)
• Code determines location and resolution
  automatically
• Adapts in all three dimensions
• The NCSU k-zeta (Enstrophy) hybrid turbulence
  model
• Four equations based on exact equations derived
  from the Navier-Stokes and modeled term by term
• MM5 has several sources of dissipation (e.g.,
  Asselin filter) that limit the resolution
• LES resolution of turbulence scales not yet
  achieved

11
11 January 1972 Boulder Windstorm

Turbulent breakdown of topographically forced
gravity waves

• 2-D test
• Same setup as in Boyle et al. (2000)

12
Adaptive Grid at t3h
13
Velocity Vectors at t3h
14
Daysmoke

• A dynamic-stochastic model consisting of
• Entraining turrets representing hot rising air
  which define the plume boundary,
• Large eddy parameterization for plume deformation
  due to turbulent fluctuations
• Detraining parcels that cross plume boundary due
  to stochastic plume turbulence
• Multiple plume boundaries can exist
  simultaneously allowing Daysmoke to simulate
  complex plume structures (e.g., multiple-core
  updrafts)

15
Daysmoke
z
x
y
x
16
Subgrid Chemistry

• Daysmoke is a non-reactive model.
• Subgrid chemistry of mean parcel concentrations
  can be modeled by tagging them as plume and the
  rest of the grid cell as ambient (Parcel-Grid
  Method of Chock and Winkler, 1994).
• Turbulent fluctuation correlations of
  concentrations can also be modeled (Advanced
  Plume Treatment of Karamchandani et al, 2000).

17
28 February 2007 Atlanta Smoke Event

• An opportunity to compare new model with base
  model in terms of agreement with
  regional observations

18
Model Evaluation Baseline
Forecast, Hindcast and Observed PM2.5
Indicator of success improved agreement of
predictions with observations
19
Summary

• In general, adaptive grid models produce more
  accurate solutions than their static grid
  counterparts with comparable or even larger
  computational demands.
• The improved versions of the models are expected
  to result in better resolved dynamics, emissions,
  and chemical transformations as well as reduced
  numerical diffusion.
• This will be checked by re-evaluation of the 28
  February 2007 smoke event in Atlanta, which has
  already been simulated by using the current
  uniform grid MM5/CMAQ modeling system.

20
Acknowledgements

• Strategic Environmental Research and Development
  Program (SERDP)
• Joint Fire Science Program (JFSP)
• Visibility Improvement State and Tribal
  Association of the Southeast (VISTAS)