NOx Emission Inversion Using the Adjoint of CMAQ

1
NOx Emission Inversion Using the Adjoint of CMAQ

• Farid Amid and Amir Hakami

Carleton University
2
Overview

• - Inversion Methodology
• OMI retrievals
• Observational operators
• Adjoint inversion
• Category-specific emission inversion
• - Results

3
Adjoint-based inversion

• Grid-based emission inference

4
Model application

• North America
• 36 km
• 13 layers
• Summer of 2007
• OMI NO2 column and surface ozone
• For now wont use surface NO2
• Category-specific as much as possible
• SAPRC-99
• Parallel (?)
• Bott scheme (?)

5
OMI retrievals

• KNMI product
• Filtered to remove
• Large pixels (gt 36 km)
• High errors (gt 70)
• Domain edge (?)
• Horizontal and vertical regridding (mapping)
• Observational operators

6
Domain regridding (forward and backward)
Averaging Kernel -- A
VCDCMAQ
AT
CMAQ forcing
VCDCMAQ- VCDOMI
7
Category-specific inversion

• Grid-based inversion does not distinguish between
  source categories when sources are collocated in
  one grid.
• Emission adjustments are only applied to the
  total emissions, i.e. adjoint gradients are
  scaled by emission shares.
• Not necessarily a problem
• It would be useful to track areas where specific
  sources dominate.

8
Source contributions to the gradients
C
A
B
B
9
Forward sensitivity analysis

• Using DDM for screening receptor sites where a
  single source category dominates (NO2 sensitivity
  wrt to anthropogenic NOx emissions)

10
Steps

• Identify sources with potentially high spatial
  correlation in emission bias.
• Natural sources, mobile (?)
• Screen for receptors where these sources
  dominate.
• Invert for those emissions.
• Extend the adjustments to other locations.
• Invert for other sources.

11
Conclusions

• Spatial correlations that are intuitive are a
  potential source of information which should not
  be neglected, particularly in absence of good
  understanding of the true covariance matrices.
• Such information can be applied to distinguish
  between sources using forward sensitivities.
• Major sources are typically collocated.
• NO2 may not be the best candidate for this type
  of analysis (biogenics?)