Skip to main content

Full text of "DTIC ADA598095: Aerosol Observability and Data Assimilation Investigations in Support of Atmospheric Composition Forecasts"

See other formats

DISTRIBUTION STATEMENT A. Approved for public release; distribution is unlimited. 

Aerosol Observability and Data Assimilation Investigations in Support of 

Atmospheric Composition Forecasts 

Jeffrey S. Reid, Walter Sessions*, Peng Lynch*, Edward J. Hyer 
Naval Research Laboratory 
7 Grace Hopper Ave., Stop 2 
Monterey, CA 93943-5502 

Phone: (831) 656-4725 Fax: (831) 656-4769 email: 

*CSC Inc., Monterey, CA 

Document Number: N0001413WX20162 


In this final year of a 3 year effort by the Naval Research Laboratory Code 7544 to enable Navy 
aerosol forecasting to take full advantage of available data feeds through the investigation of 
fundamental atmospheric composition observability, we find that the next generation of data feeds and 
associated technology present a number of challenges and opportunities which require attention. These 
include: 1) The transition from NASA EOS sensors to the next generation of diversified operational 
and near real time data sources; 2) The move to a constellation approach for global atmospheric 
composition observing, 3) The expected near real time availability of US and European lidar data; 4) 
The enhanced availability of surface and aircraft observations; and 5) Increased aerosol model 
demands for such applications as joint surface-atmosphere retrievals, directed energy (DE), and 
intelligence, surveillance, and reconnaissance (ISR). This increase in the number of potential data 
sources, coupled with further efficacy demands, creates an imperative need to understand the nature of 
global aerosol observability and the development of realistic uncertainties for composition 
observations, retrievals and model products. Outstanding problems facing the community relate to 
such issues as observation bias, representativeness, and information spreading for the myriad of 
sporadic data sources available. To meet this need, in this grant we investigated the use of these 
diverse flows of data using ensemble and ensemble data assimilation technologies to be incorporated 
into the Navy Aerosol Analysis and Prediction System (NAAPS)/Navy Variational Analysis Data 
Assimilation-Aerosol Optical Depth (NAVDAS-AOD) framework. This work also led to the 
development of the world’s first quasi-operational global multi-model ensemble, drawing from the 
world’s operational data centers. This is the nucleus of a potential operational multi-model ensemble. 
As with other consensus-like products, such an ensemble is likely to remain top ranked. 


Our overarching goal is to investigate applied science aerosol observability issues related to the proper 
determination of observed product efficacy and information spreading. To this end the core 
methodology involves the development and application of an ensemble version of NAAPS and its 
subsequent ensemble based data assimilation system. At the same time, we have initiated the 


Report Documentation Page 

Form Approved 

OMB No. 0704-0188 

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and 
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, 
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington 

VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it 
does not display a currently valid OMB control number. 


30 SEP 2013 2 ' REPORT TYPE 


00-00-2013 to 00-00-2013 


Aerosol Observability and Data Assimilation Investigations in Support of 
Atmospheric Composition Forecasts 









Naval Research Laboratory,? Grace Hopper 








Approved for public release; distribution unlimited 






18. NUMBER 19a. NAME OF 



unclassified unclassified unclassified Report (SAR) 


Standard Form 298 (Rev. 8-98) 

Prescribed by ANSI Std Z39-18 

development of a multi-model ensemble across major centers including European ECMWF, Japanese 
JMA, NASA GMAO, NOAA NCEP and the Spanish BSC. Subcomponents of our effort include: 

1) The development of an ensemble-based NAAPS system which makes use of the FNMOC 
NOGAPS ensemble data set: Here we will use the base NAAPS model, including the data 
assimilation system and run in parallel with the 20 NOGAPS fields, further perturbed with 10 
to 30 additional perturbations to source and sink functions. This version of NAAPS will be 
prepared for eventual transition to 6.4 and FNMOC if the Navy so desires. 

2) The development of key verification metrics of models and observations alike: Based on initial 
simulation of the ensemble, we will investigate appropriate error metrics for Navy satellite and 
model products that account for the non-Gaussian, spatially correlated nature of biases present 
in the environment. This objective includes the development of probabilistic tools for forecasts. 
These can then be applied to satellite observations for model verification. 

3) Investigation into the nature of available satellite and in situ observations including their flow 
dependant correlation lengths for use in NAVDAS-AOD: The optimal mix of meteorological 
and source/sink perturbations for the NAAPS ensemble for given computational constraints 
will be identified. A particular emphasis will be on the application of these methods to aerosol 
vertical distribution. 

4) The incorporation of NAAPS in the NCAR Data Assimilation Test-bed (DART): Once the 
NAAPS ensemble model is running quasi-operationally and is well understood, we will port 
NAAPS to the DART package at NCAR. This is the first step in developing our ensemble- 
based data assimilation system. 

5) The creation of a research grade ensemble-based data assimilation system: Based on our 
experience with the DART system, we anticipate being able to have the ability to harvest and 
develop a robust ensemble based data assimilation system. This will be used to supplement the 
existing 3DVAR aerosol data assimilation system currently in use. Emphasis in this system 
will be on sparse or isolated observations such as provided by lidars, sun photometers and 
surface reports. In addition, ensemble co-variability studies will be carried out to address 
phenomenological questions such as modes of atmosphere/aerosol coupling. 

6) Collaborate with other centers to derive an aerosol consensus for severe events: It has been 
shown that for many extreme events, consensus forecasting is a top perfonner. In this objective 
we wish to investigate the applicability of performing multi-model consensus of GMAO, 
ECMWF and other centers to investigate the predictability and extent of significant aerosol 


As outlined in previous reports, in FY12 and 13, we developed the basic architecture of both the 
NAAPS single and ICAP-multi-model ensembles. Work perfonned in FY13 focuses on the application 
of these systems for better understanding aerosol observability. The spatial covariance fields are 
generated in an ensemble version of NAAPS (henceforth eNAAPS). Nominally eNAAPS is run quasi- 
operationally using the 20 NOGAPS ensemble members generated by FNMOC every 12 hours. 
Additionally, other members, based on perturbations to the NAAPS source function, are created. 


Results from the initial stages of analysis led to a series of upgrades for eNAAPS sources and 
microphysics (such as a new pollution specie to replace sulfate) which are available for the 
detenninistic operational version once the NAVGEM porting is complete. 

Collinear with the development of eNAAPS has been the continued development of the multi-model 
ensemble. Based on members of the International Cooperative for Aerosol Prediction (ICAP), the 
ICAP Multi-model ensemble (or ICAP-MME) is the world’s first quasi-operational aerosol ensemble, 
and is already seeing significant use across centers. There are four multi species models is ICAP- 
next generation. In addition to these, for the dust species only, we also incorporate NOAA NCEP’s 
NGAC and the Barcelona Supercomputing dust models. Data for all models now goes back 18 
months, with the time series for the core models going back 28 months. 

Tools and display systems were also developed for analysis of ensemble products. Basic tools for 
analyzing ensemble variably, correlation length scales, and the most commonly employed “spaghetti 
plots” and ra nk histograms were largely adapted from existing Matlab code into Perl, Python and IDL 
framework currently used in 7544 and FNMOC. Online verification products based on AERONET, 
satellite data and own analyses are being developed. In a similar manner to eNAAPS, we cooperate 
with other meteorological centers over the globe which are developing global aerosol prediction 
capabilities. By pulling data from such centers as ECMWF, JMA, NASA GMAO, and NOAA NCEP, 
we created a first of its kind global aerosol multi-model ensemble. For comparison purposes, products 
generated are similar in nature to eNAAPS. 

For ensemble based data assimilation we ported NAAPS into the NCAR Data Assimilation Research 
Testbed (DART) in years one and two. DART hosts numerous ensemble-based data assimilation tools 
including the core components of an EnKF system. Both NOGAPS and COAMPS have already been 
ported to DART, and ensemble based COAMPS dust source functions studies are already underway 
through a joint project with Dr. Hansen. Hence, in house knowledge already exists to expedite this 
process. Included in this budget is additional travel money so that NRL developers can spend the 
required time at NCAR to make this port happen. 

Finally, additional verification tools need to be generated to evaluate improvements to the model. This 
not only allows model parameterizations, but also the satellite data assimilated into the models to be 
investigated and improved. In particular, technologies need to be created that can quickly evaluate 
new or updated satellite products, such as those being generated by VIIRS. 


In FY 13, the following tasks were completed: 

1) eNAAPS was fully exercised. Most importantly, in conjunction with Dean Hegg (University of 
Washington) we ported the current NAAPS sulfate model to a more generic “pollution” specie 
to account for the more dominant secondary organic aerosol loadings in the atmosphere. Once 
NAVGEM development and tuning is complete, we feel the NAAPS ensemble is ready for 
transition to 6.4. 

2) We performed an extensive analysis of the ICAP multi-model ensemble for a one year period 
of perfonnance. We rigorously compared ICAP-MEE and its single model components against 


a series of continuous (e.g., RMSE, MAE etc.) and threat score based metrics. Regions of 
common success and difficulty over the globe were mapped. 

3) Publications on eNAAPS and ICAP-MME were initiated, and at the end of the fiscal year 
rough drafts of findings will be complete 


Research results for this project can be divided into areas of ensemble development, verification, and 
data assimilation. In FY11 & 12, focus was on development and data assimilation. For FY13, focus has 
been on verification and analysis, with additional effort on the creation of a pollution specie in 
NAAPS, and the tuning of the dust and smoke parameters. 

Pollution development and tuning: Figure 1 demonstrates before and optimized versions of eNAAPS, 
based on NOGAPS deterministic meteorology. The largest single improvement was due to the 
incorporation of secondary organic aerosol mass into the sulfate specie, creating a “fine mode pollution 
specie.” Areas of consistent difficulty for NAAPS including East Asia, the Indian sub-continent, and 
the eastern United States, have been dramatically improved. The mechanism uses maps of isoprene, 
terpene and aromatic emissions, largely from the MACC and MEGAN source functions, to develop a 
new source for fine mode particles. This source is not static, but is based on environmental conditions 
and the pre-existing aerosol load. Thus, this formulation does not technically add any additional 
aerosol or gas species to be carried in the model. In addition to this, based on the previous ensemble 
Kalman filter work, the dust source map and biomass burning emission have been improved. In Asia, 
the Indian sub-continent and sub-Saharan Africa, RMSE values were improved by 40%. Mean 
innovations to account for the lacking source function for fine mode organic aerosol particles were also 
dramatically reduced. 

Ensemble development and analysis: In year three we examined a number of case studies of severe 
dust and biomass burning events. An example of this is depicted in Figure 2, where both the eNAAPS 
“spaghetti plot” and several other ICAP-MME members are presented. In all of these cases, while a 
dust front was predicted by all model members, magnitudes differed greatly. Diversity in the models 
could be tracked back to a dust source in the Taklimakan desert. Aerosol data assimilation in and of 
itself does not help in this situation, as the dust is generated in a desert area where aerosol optical 
thickness retrievals are rare and inaccurate. Further, during transport, dust was masked by a frontal 
cirrus shield. However, using the NRL dust enhancement product we could determine semi- 
quantitatively which ensemble members were performing correctly. In an operational setting, 
forecasters could use the ensemble in this way with Navy imagery to accurately forecast a wave of dust 
passing over the Korean Peninsula and the Sea of Japan. 

Verification: Both eNAAPS and ICAP MME underwent extensive verification and analysis in FY13. 
From this, a common climatology was generated and areas of common difficulty among models were 
identified. Examples of both these aspects are presented in Figure 3. In the upper set of panels, total, 
fine, and coarse mode AOT medians and geometric standard deviations are presented for the ICAP- 
MME ensemble average. Inset as dots on this figure are AERONET verification values. Overall, we 
find that all models from all centers tend to underestimate AOT globally against AERONET. Further, 
while all models reproduce major aerosol features (e.g., dust plumes in Africa and Asia, major biomass 
burning regions, etc.), there are at times very large differences in mean amplitude and minimum AOT. 
Particularly large differences exist in in the mid-latitude oceans, presumably due to differences in sea 


salt source functions. In both cases, these are important findings that will lead to future work to further 
improve aerosols source/sink functions and data assimilation. 


We expect much of this work to be of immediate use to the warfighter. Just as the current NRL 
aerosol page is frequently used in the METOC community, we expect these ensemble products to be of 
immediate applicability. To begin with, in key portions of the globe we can generate dust, smoke, or 
pollution “Spaghetti plots” for each of the meteorological members (Figure 4) or independent models. 
These can also be shown spatially as areas of high variability (Figure 2). In these areas we will 
collaborate with Dr. Hansen and Dr. Whitcomb who are working on METOC impacts, decision aids, 
and scorecards. 


We have begun discussions with FNMOC to transition eNAAPS to operations. We are currently 
waiting for the next set of revisions of NAVGEM to take place. Once the NOGAPS ensemble 
transitions to the NAVGEM ensemble, we anticipate a final round of model optimization to take place 
before we initiate 6.4 work. 


This project is tightly coupled to a number of ONR 32 programs, particularly those of Professor 
Jianglong Zhang at the University of North Dakota. Our primary transition partner is Douglas 
Westphal, who is principal investigator on the Large-Scale Aerosol Model Development (PI: Doug 
Westphal). New data-processing and visualization systems are being adapted for aerosol research 
through the COAMPS-On Scene (COAMPS-OS Cr ) 1 IVPS charts program (PI: John Cook). We have 
also begun working with Jim Hansen on his ONR-funded project for the use of ensemble data 
assimilation in the prediction of atmospheric constituents. 


Journal Publications 

Toth, T. D., J. Zhang, J. Campbell., J. S. Reid, et ah, 2013, Investigating elevated Aqua MODIS 

aerosol optical depth retrievals over the mid-latitude Southern Oceans through intercomparison 
with co-located CALIOP, MAN, and AERONET datasets, J. Geophys. Res, in press 

Eck, T. F., B. N. Holben, J. S. Reid, et ah, (2013), A seasonal trend of single scattering albedo in 
southern African biomass burning particles: Implications for satellite products and estimates of 
emissions for the world's largest biomass burning source, J. Geophys. Res., 118, 

Johnson, R. S., Zhang, J., Hyer, E. J., Miller, S. D., and Reid, J. S. (2013), Preliminary investigations 
toward nighttime aerosol optical depth retrievals from the VIIRS Day/Night Band, Atmos. Meas. 
Tech., 6, 1245-1255, doi:10.5194/amt-6-1245-2013. 

COAMPS-OS® is a registered trademark of the Naval Research Laboratory. 


Khade, V. M., J. A. Hansen, J. S. Reid , and D. L. Westphal, (2013), Ensemble filter based estimation 
of spatially distributed parameters in a mesoscale dust model: experiments with simulated and 
real data, Atmos. Chem. Phys., 13, 3481-3500, doi:10.5194/acp-13-3481-2013. 

Shi, Y., Zhang, J., Reid, J. S ., Hyer, E. J., and Hsu, N. C., (2013), Critical evaluation of the MODIS 
Deep Blue aerosol optical depth product for data assimilation over North Africa, Atmos. Meas. 
Tech. 6, 949-969, doi:10.5194/amt-6-949-2013, 2013. 

Other Publications 

Eck, T. F., B. N. Holben, J. Schafer; T. Berkoff, J. S. Reid , et ah, (2012) Observed enhancements in 
aerosol optical depth in the vicinity of cumulus clouds during DISCOVER-AQ , American 
Geophysical Union Fall Meeting, Dec. 3-7 San Francisco, CA, A13K-0328. 

Geiszler, D. A., J. Cook, S. Chen, M. Frost, P. Harasti, C. Hutchins, Q. Zhao, J. S. Reid, D. Martinez, 
R. A. Allard, T. J. Campbell, T. A. Smith, and L. McDermid, (2013), COAMPS-OS®: An 
atmospheric/ocean/wave prediction system for the coastal environment. American meteorological 
Society Annual Meeting, January 5-12, Austin TX. 

Hyer, H. J„ J. S. Reid , J. Zhang, C. A. Curtis, (2013), Satellite aerosol observations for air quality: 
matching the scales of observations and applications, American Geophysical Union Fall Meeting, 
Dec. 3-7 San Francisco, CA, A11L-01. 

Hyer, E. J„ J. Zhang, J. S. Reid , W. R. Sessions. C. A. Curtis, and D. L. Westphal, (2013), Operational 
assimilation of aerosol optical depth from NPP VIIRS in a global aerosol model, American 
meteorological Society Annual Meeting, January 5-12, Austin TX. 

Johnson, R. S., J. Zhang, J. S. Reid, E. J. Hyer, S. D. Miller, (2012), Impact of VIIRS on Multi-Sensor 
Aerosol Optical Depth Data Assimilation, American Geophysical Union Fall Meeting, Dec. 3-7 
San Francisco, CA, A13J-0318. 

Shi, X., J. Zhang, J. S. Reid , B. Liu, R., Deshmukh, (2013), Critical evaluation of cloud contamination 
in MISR aerosol product using collocated MODIS aerosol and cloud products, American 
Geophysical Union Fall Meeting, Dec. 3-7 San Francisco, CA, A13J-0310. 

Zhang, J., J. S. Reid, J. R. Campbell, E. J. Hyer , R. S. Johnson, Y. Shi, and D. L. Westphal, (2013), 
Sensitivity of aerosol climate forcing to the existing satellite observations, American 
Meteorological Society Annual Meeting, January 5-12, Austin TX. 



- 0.1 - 0.01 0 0.01 0.1 

Mean AOT Innovation (550 nm) 

2011 Natural RunAOTs 

0.5 0.7 1.0 1.2 

Aerosol Optical Thickness (550 nm) 

AOT Corrections 

Figure 1. Original and optimized Aerosol Optical Thickness (AOT) and mean AOT innovation with 
the original NAAPS, and an optimized version which included the incorporation of secondary 

organic aerosol in the model. 


Figure 2. An example of differences within eNAAPS and the ICAP Multi-Model Ensemble (ICAP- 
MME) for a single 72 hour dust event in November 2012. 


Coarse Fine Total Coarse Fine Total 





Median AOT (550 nm) 

AOT Geometric Standard Deviation 


Maximum Minimum 


Maximum Minimum 

0 0.04 0.08 0.16 0.32 0.64 1.28 

Pointwise AOT Maximum and Minimum (550 nm) 

Figure 3. Bi-seasonal sets of statistics for the ICAP Multi-model ensemble for a 1 year test period. 
Upper set: the seasonal ICAP-MME median and geometric standard deviation of total, fine and 
coarse mode AOT. Lower set: a point-wise comparison of the maximum and minimum mean aerosol 
optical thickness (AOT) drawn from the ensemble members. That is, each point on the map draws 
from the average maximum and minimum from the ensemble members, thus showing the maximum 
average or minimum average AOT for any point. Shown are total, fine, and coarse mode AOT. 

Inset are averages derived from AERO NET