Progress Report

2004 Boulder Meeting

Other Messages

Problems to Address in NSF ITR John H. Seinfeld Oct. 10, 2002
Aerosol assimilation in MOZART/MATCH Peter Hess Oct. 11, 2002
Advancing the integration of observations and CTM's Tad Anderson and Tami Bond Oct. 14, 2002
Things needed before starting Greg Carmichael January, 2003

Problems to Address in NSF ITR

John H. Seinfeld
We have chosen to focus on inverse problems, data assimilation, and sensitivity analysis in regional and global aerosol chemical transport models (CTMs). As we discussed, the initial phase of the project will be to derive the adjoint model corresponding to the STEM/CFORS model and develop the necessary numerical analysis to implement the model. Subsequently, we will consider extension to a global aerosol CTM, perhaps through collaboration with NCAR.

The purpose of this note is to propose scientific problems that can be addressed with the models. I will focus, at this point, at the regional (sub-hemispheric) scale, which, practically, could be East Asia or North America.

1. Inverse Problem - Determination of the Spatial Distribution and Strength of Black Carbon (BC) emissions that produce the optimal fit of predicted and observed BC concentrations at an array of surface sites. This problem can be viewed analogously to that of the determination of optimal CO emissions in a gas-phase CTM, for which several recent papers exist. This is a good starting problem because BC is a non-reactive, conservative tracer, so issues of aerosol microphysics need not be dealt with.

2. Inverse Problem - Same as 1 for Organic Carbon (OC). Determine primary OC emissions inventory that produces optimal fit of predicted and measured OC concentrations for same region as in 1. This is slightly more complicated than 1 since measured OC may include a contribution from secondary OC, which is not included in the OC emissions inventory. The degree of mismatch of the concentrations at the optimal fit can be an indication of the amount of secondary OC in the measured data. For both 1 and 2, it will be necessary to identify regions that contain sufficient measurements of BC and/or OC.

3. Sensitivity Analysis - Determine the sensitivity of predicted aerosol concentrations to uncertainties (i.e. variations) in emissions and atmospheric conditions: 

                 SO2 emissions           ---    sulfate concentrations
                 cloud perturbations     ---    sulfate concentrations
                 temperature             ---    nitrate concentrations
                 NOx emissions           ---    nitrate concentrations
(The cloud perturbation scenario may not be easy to do, but I list it as a possibility, as it would assess the amount of sulfate being formed by in-cloud processes. Sensitivity coefficients can be calculated from the adjoint solution.)

4. Assimilation of Remote Sensing (Satellite) Data - Satellite data can be used to constrain the integrated burden of aerosol, via its radiative signal, at certain times and locations. The aerosol CTM predicts those column burdens based on our best a priori information on emissions and aerosol properties. Either emissions or atmospheric aerosol properties can be varied to produce an optimal fit of predicted and observed column burdens. This is essentially an inverse problem like 1 and 2, but with a different type of measurement.

I think a reasonable goal is to attack these problems with STEM/CFORS over a particular region within the first two years of the project.

Aerosol assimilation in MOZART/MATC

Peter Hess

There are currently a large number of satellites with aerosol sensors. Some examples:
Each of these satellites has somewhat different characteristics......

I think there are lots of advantages in using satellite data on a global scale:

Activities at NCAR:

Models:

The MOZART model does have an aerosol module. It includes global emissions for the primary aerosol types. The strength of MOZART is its chemistry. It is also highly optimized and simulates 1 year in about a day at moderate resolution (T42) with over 50 species (and over 150 reactions) on the IBM here at NCAR. This speed will increase. The chemistry module is highly flexible so the number of species can easily be changed. The weakness is the aerosol microphysics. In the next year, MOZART, MATCH and the CCSM (Community Climate System Model) should all be running in one framework.

An adjoint to the MOZART transport model has not been fully developed, but there are some efforts in this direction.

There are many ways one could go, but a reasonable start might be to do a global data assimilation around the time of ACE-Asia.

The following steps might be considered in the future.

Advancing the integration of observations and CTM's

Tad Anderson and Tami Bond

The goal of the ITR project is to advance the art of the CTM, specifically, by figuring out how to achieve better integration with observations. This means making better use of existing observations and designing better observational strategies. We envision a role for aerosol variability in this effort.

We come to this ITR project as an aerosol measurement group. We are interested in developing analysis tools for measurements that either parallel or complement state-of-the-art transport models. Complementary approaches include smoothing or otherwise 'massaging' existing observations to facilitate assimilation into models. Parallel approaches include identifying features of four-dimensional chemical and optical fields that should be common between measurements and models, such as higher moments of the data or spatial and temporal variability.

1. Aerosol variability - general discussion

Most model/obs. comparisons are framed in terms of average concentration (Langner and Rodhe, 1991; Benkovitz et al., 1994; Benkovitz and Schwartz, 1997) or average Aerosol Optical Depth (Ghan et al., 2001). Some comparisons have considered intensive properties like single scattering albedo (Liousse et al., 1996; Heintzenberg et al., 1997; Ghan et al., 2001). But very few have explicitly considered variability.

Yet the time/space scales of variability are closely linked to the physical processes that control concentration. Moreover, correlations among variables (at various time/space scales) can provide further insights into causes. Thus, model/obs comparisons that are focused on variability may have enhanced diagnostic value over simple comparisons of average properties.

This will be a largely exploratory aspect of the larger ITR project. The key challenges will be:
  1. identify and obtain aerosol data sets suitable for variability analysis and work these into common formats for analysis. (This requires learning about the strengths and limitations of each data set and usually doing some filtering.)
  2. identify and adapt appropriate statistical tools for analyzing variability. (A key question here is how to distinguish ambient variability from instrumental noise and artifacts)
  3. develop methods of comparing variability between models and observations. (What parameters are suitable for comparison? Do the observations need to be smoothed to the resolution of the models? How do we interpret discrepancies?)
Luckily, our group has separate funding through NSF to undertake steps i. and ii. above. Thus, we can concentrate on iii for this project. Some preliminary thoughts and possible directions...
2. Variability analysis of ACE-Asia intensive campaign and design of a long-term monitoring strategy
As a specific activity, we propose multivariate, factor analysis of some significant portion of the data from the ACE-Asia campaign. For example, a merged data set already exists for the C-130 aircraft measurements and includes aerosol optics, size, chemistry, trace-gas concentrations (ozone, CO, hydrocarbons, etc), and ambient meteorology. The idea would be to identify a small number of factors that account for most of the overall variability then
  1. examine the models to see if these same correlations exist
  2. if not, try to diagnose why and make adjustments
  3. as a joint model/obs activity, try to design a practical strategy for low-cost, long-term measurements in the Asian outflow region that would capture the fundamental modes of variation identified during the intensive and would provide the optimal, observational constraint for models in the future. (Caveat: This activity will have to look beyond just the data from the Spring, 2001 intensive, since that fails to capture the annual cycle.)
3. General comment on group plans to date (from Tad)
It is good to be exploring a wide spectrum of ideas at this point in the game. The next step will be to flesh them out with specific tasks, identified data sets, and careful assessments of data quality. Many of these ideas will fail to pan out due to insufficient data or inadequate data quality. The recent paper by Prasad Kasibhatla et al (2002) is instructive. They successfully used a simple inversion method to identify problems with regional-scale CO emission inventories. But they weren't able to infer anything very detailed about the emissions and it seems to me that CO is MUCH more amenable to this sort of analysis than aerosols. Why? CO can be readily and unambiguously measured, the concentration data is fairly widespread and longterm (and standardized), CO has relatively simple chemistry, and it has a very convenient atmospheric lifetime of about 30 days. We should not just assume that a given aerosol species or property will be amenable to inverse modeling; rather, we will have to pick our aerosol targets carefully. To be frank, BC and OC make me nervous. Although I realize there is tremendous interest in these species, I'm not sure there is a sufficient quantity of sufficiently accurate data to constrain an inversion. (Of course, the matter needs to be decided not by intuition but by a proper propagation of uncertainties.)
References cited
  1. Benkovitz, C. M., C. M. Berkowitz, et al. (1994). "Sulfate over the North Atlantic and adjacent continental regions: Evaluation for October and November 1986 using a three-dimensional model driven by observation-derived meteorology." J. Geophys. Res. 99: 20725-20756.
  2. Benkovitz, C. M. and S. E. Schwartz (1997). "Evaluation of modeled sulfate and SO2 over North America and Europe for four seasonal months in 1986-1987." J. Geophys. Res. 102: 25305-25338.
  3. Davis, A., A. Marshak, et al. (1996). "Scale invariance of liquid water distribution in marine stratocumulus. Part I: Spectral properties and stationarity issues." J. Atmos. Sci. 53: 1538- 1558.
  4. Ghan, SF, Laulainen, N, et al. (2001) "Evaluation of aerosol direct radiative forcing in MIRAGE," J Geophys Res, 106, 5295-5316.
  5. Heintzenberg, J., R. J. Charlson, et al. (1997). "Measurement and modelling of aerosol single-scattering albedo: Progress, problems and prospects." Beitr. Phys. Atmosph. 70: 249-263.
  6. Kasibhatla, P. et al. (2002). "Top-down estimate of a large source of atmospheric carbon monoxide associated with fuel combustion in Asia," Geophys. Res. Lett., 29(19), 1020/2002GL015581.
  7. Langner, J. and H. Rodhe (1991). "A global three-dimensional model of the tropospheric sulfur cycle." J. Atmos. Chem. 13: 225-263.
  8. Liousse, C., J. E. Penner, et al. (1996). "A global three- dimensional model study of carbonaceous aerosols." J. Geophys. Res. 101: 19411-19432
  9. Ott, W. R. (1990). "A physical explanation of the lognormality of pollutant concentrations." J. Air Waste Manage. Assoc. 40: 1378-1383

Things needed before starting

Greg Carmichael

  1. Decide which model(s) to use
  2. We need a very "clean" implementation/documentation of the transport-chemistry model. Sandu will look into STEM.
  3. We must have a good undeerstanding of the relationship (mapping) between the observational data and the model state.

Revised: 7/15/04 by Tianfeng Chai