March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
CHAPTER 16
Data Assimilation and Objectively Optimized Earth Observation
David J. Lary and Anuradha Koratkar
Global Modelling and Assimilation Office, NASA Goddard Space Flight Center, MD, USA
GEST at the University of Maryland Baltimore County, MD, USA
E-mail: David.Lary@umbc.edu
This chapter describes a vision for a future objectively optimized earth obser-
vation system with integrated scientific analysis. The system envisioned will dy-
namically adapt the what, where, and when of the observations made in an online
fashion to maximize information content, minimize uncertainty in characterizing
the systems state vector, and minimize both the required storage and data pro-
cessing time for a given observation capability. Higher level goals could also be
specified such as the remote identification of sites of likely malaria outbreaks.
By facilitating the early identification of potential breeding sites of major vector
species before a disease outbreak occurs and identifying the locations for lar-
vicide and insecticide applications. This would reduce costs, lessen the chance
of developing pesticide resistance, and minimize the damage to the environ-
ment. Here we describe a prototype system applied to atmospheric chemistry
with two relatively mature symbiotic components that seeks to achieve this goal.
One component is the science goal monitor (SGM), the other is an Automatic
code generation system for chemical modeling and assimilation (AutoChem) de-
scribed online at http://gest.umbc.edu/AutoChem/. The Science Goal
Monitor (SGM) is a prototype software tool to determine the best strategies for
implementing science goal driven automation in missions. The tools being de-
veloped in SGM improve the ability to monitor and react to the changing sta-
tus of scientific events. The SGM system enables scientists to specify what to
look for and how to react in descriptive rather than technical terms. The system
monitors streams of science data to identify occurrences of key events previ-
ously specified by the scientist. When an event occurs, the system autonomously
coordinates the execution of the scientist’s desired goals. The data assimilation
system can feed multivariate objective measures to the SGM such as informa-
tion content and system uncertainty so that SGM can schedule suitable obser-
vations given the observing system constraints. The observing system may of
course be a sensor web suite of assets including orbital and suborbital platforms.
Once the observations are made an integrated scientific analysis is performed
1

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
2 D.J. Lary
which automatically produces a cross-linked web site for easy dissemination
and to facilitate investigation of the scientific issues. A prototype is available
at http://gest.umbc.edu/CDACentral/.
Keywords: Optimized Earth Observation, Science Goal Monitor, Data Assimi-
lation
1. Introduction
This year, 2004, is the fortieth anniversary of the NASA Nimbus program. The
Nimbus satellites, first launched in 1964, carried a number of instruments: mi-
crowave radiometers, atmospheric sounders, ozone mappers, the Coastal Zone
Color Scanner (CZCS), infrared radiometers, etc. Nimbus-7, the last in the series,
provided significant global data on sea-ice coverage, atmospheric temperature, at-
mospheric chemistry (i.e. ozone distribution), the Earth’s radiation budget, and
sea-surface temperature.
What will the earth observing systems of the future look like? Will they be au-
tonomous? This chapter describes one vision for future earth observing systems.
New in this vision is the desire for symbiotic communication to dynamically guide
an earth observation system. An earth observation system which is not just a sin-
gle satellite acting on its own but a constellation of satellites, and sub-orbital plat-
forms such as unmanned aerial vehicles ?;?;? (http://uav.wff.nasa.gov/),
and ground observations interacting with computer systems used for model-
ing, data analysis and dynamic observation guidance. Automatic code genera-
tion (http://gest.umbc.edu/AutoChem/) and automatic parallelization will
greatly facilitate the implementation and automatic adaption of the system for
different problems and its possible use on a variety of hardware. Automatic doc-
umentation of both software and data products facilitate both code maintenance,
and the production and quality monitoring of self-consistent analyses. These anal-
yses can be used by scientists to understand and answer major scientific questions,
and by policy makers to establish sound policy decisions, thus increasing the ac-
cessibility and utility of Earth Science data. Automatic compression minimizes
both the required cost of storage and dissemination, and the required time for
electronic product transfer/download.
Most of the key questions in earth science involve the tracking of dynamically
evolving geophysical fields. So it is desirable to make the best use of a given
earth observation capability by using an objective dynamic data retrieval control
system that dynamically adapts the observations made in an online fashion. This
facilitates the dynamic tracking of time-evolving sharp gradients, one example
would be those in chemical tracer fields often located at the polar vortex edge, the

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 3
tropopause and the day-night division. An example of this is shown in Figure ??.
On the left the sharp gradient in NO (nitric-oxide) at the terminator can be seen.
On the right a visualization of both the tropopause and the polar vortex can be
seen, both are important mixing barriers.
This approach fits in
well with the Sensor Web Concept (http://sensorwebs.jpl.nasa.gov/).
A Sensor Web consists of a group of sensors (satellites, UAVs, aircraft,
ground stations) set up to collect various kinds of information, communicate
with other sensors in the web. NASA is already taking the first steps toward
Internet-like connectivity among its Earth sensing satellites. A system com-
posed of multiple science instrument/processor platforms that are interconnected
by means of a communications fabric for the purpose of collecting measure-
ments and processing data for Earth or Space Science objectives. An exam-
ple how these ideas have already been used to track fires is available online at
http://earthobservatory.nasa.gov/Newsroom/NasaNews/2003/
2003072215047.html. Imagine, for example, if all the NASA earth science
satellites currently in orbit (Figure ??) had active communication with each other
and with an intelligent observation direction system and global modeling tools.
Such an arrangement would allow a much greater synergy than is currently possi-
ble and thus allow for an objectively optimized approach to earth observation.
NASA and ESA science missions have traditionally operated on the assump-
tion that we can only manage scheduling priorities and scientific processing on
the ground with significant human interaction, and that all scientific data must be
downloaded and archived regardless of its scientific value. However, increases in
onboard processing and storage capabilities of spacecraft, as well as increases in
rates of data accumulation will soon force NASA operations staff and scientists
to re-evaluate the assumption that all science must be done on the ground. In or-
der to take advantage of these new in-flight capabilities, improve science return
and contain costs, we must develop strategies that will help reduce the perceived
risk associated with increased use of automation in all aspects of spacecraft op-
erations. An important aspect of science operations is the ability to respond to
science driven events in a timely manner. For such investigations, we must teach
our observing platforms to intelligently achieve the scientists goals. The principles
presented here are generic but a specific example will be taken from atmospheric
chemistry. The assimilation system described is being used in the NASA Global
Modeling and Assimilation Office to assist with EOS Aura data validation.
Throughout this chapter a central concept is that of state vectors. The first
step in the mathematical formalisation of the system is the definition of the work
space. The collection of numbers needed to represent the state of the system being

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
4 D.J. Lary
studied is collected as a column matrix called the state vector, x.
2. Dynamic Data
The first and most important element is the concept of dynamic data . The dynamic
data retrieval control system envisioned here dynamically adapts what measure-
ments are made, where they are made, and when they are made. The dynamic
adaption is performed online to maximize the information content, minimize the
uncertainty in characterizing the systems state vector, and minimize both the re-
quired storage and data processing time, and minimize the data heterogeneity min-
imization within analysis grid cells. (It is conceivable that these ideas could be
used in future to direct additional observations from unmanned automated sub-
orbital platforms.)
The ability to develop a dynamic data retrieval control system for an objec-
tively optimized earth observation system depends in large part on products made
available when data assimilation is an integrated part of the earth observation sys-
tem. Making data assimilation an integral part of the earth observation system is a
prudent step since assimilation seeks to bring together heterogeneous information
together with its associated uncertainty from a variety of sources (both observa-
tional and theoretical) in a self-consistent mathematical framework.
3. Science Goal Monitor
At the heart of the dynamic data retrieval control system is the Science Goal Mon-
itor. The Science Goal Monitor (SGM; http://aaa.gsfc.nasa.gov/SGM) is a
prototype software tool to explore strategies for implementing science goal driven
operations for multiple sensors/platforms ? . A space science SGM is being proto-
typed for dynamic automated reactions to intrinsically varying astronomical phe-
nomenon using one of the Small and Moderate Aperture Research Telescope Sys-
tem (SMARTS) telescopes. An earth science prototype has been built for Earth
Observing 1 (EO-1) to evaluate how multiple sensors can react dynamically to
obtain rapid observations of evolving earth science events. Here we envision ex-
tending these previous prototypes to use objective metrics such as information
content and system uncertainty so that SGM can schedule suitable observations
that objectively optimize the use of our assets.
Higher level goals could also be specified. For example, malaria is a major in-
ternational public health problem, causing 300-500 million infections worldwide
and approximately 1 million deaths annually. If we have developed a risk model
to predict the occurrence of malaria and its transmission intensity and its mapping
to satellite-derived and meteorological data we could ensure that our earth observ-

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 5
ing system makes observing such conditions a priority. This would then facilitate
the early identification of potential breeding sites of major vector species before
a disease outbreak occurs and identify the locations for larvicide and insecticide
applications in order to reduce costs, lessen the chance of developing pesticide
resistance, and minimize the damage to the environment. Such projects already
exist, for example, the NASA healthy planet project on Mekong Malaria and Fi-
lariasis, http://healthyplanet.gsfc.nasa.gov/project3.html.
4. Information Content and State Vector Uncertainty
As a dynamic system evolves with time not all of the state variables within the
state vector contain equal amounts of information (information content), and not
all state variables are known to the same precision. It is therefore clearly desirable
that the observations made both contain the maximum information content possi-
ble with a given observing platform capability and allow the systems state to be
characterized with a minimum uncertainty.
Information content is a broad term that could be quantified in any number
of ways depending on the system or problem being studied. Therefore, although
we propose to use a specific measure of information content for the atmospheric
chemistry system, these measures could easily be substituted with alternative mea-
sures that may be more suitable depending on the given objectives of an investi-
gation . Although we describe a specific example from atmospheric chemistry, the
principle is clearly more general. The key new concept in this approach is that
information content and system uncertainty are used in determining: what should
be measured, when and where, thus providing a cost effective strategy for using
resources and minimizing the data storage required to characterize a system with
a given level of precision.
One measure of information content/ranking that could be used is described
by ? coupled with the so-called goal attainment algorithm to provide the infor-
mation content ranking. The chemical assimilation system will provide analyses
of the state vector together with an associated uncertainty. The information con-
tent/ranking software uses the analyzed state vector to provide the information
content ranking. This information is then passed to the SGM to allow it to objec-
tively determine the following days observation schedule.
5. Automatic Code Generation
The complexity of atmospheric chemistry varies tremendously with location:
From the relatively simple chemistry of the mesosphere involving primarily oxy-
gen, hydrogen, and nitrogen containing species. To the more complex chemistry

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
6 D.J. Lary
of the stratosphere also involving chlorine, bromine, iodine, and sulphur contain-
ing species and simple hydrocarbons such as CH4 and other greenhouse gasses.
To the very complex chemistry of the troposphere, which also involves volatile or-
ganic hydrocarbons (VOCs) and their host of oxidation products. Therefore, any
tool that is going to be involved in implementing a dynamic objectively optimized
earth observation strategy must be capable of dealing with these very different
chemical regimes. Consequently, it is most desirable to have an automatic code
generator that is capable of creating and reusing code for the deterministic mod-
els required to describe the chemistry of these different regimes together with the
entire data assimilation infrastructure required (i.e. time derivatives, Jacobians,
Hessians, adjoints, and information content). The AutoChem code generation and
modeling/assimilation system has these capabilities and has already been vali-
dated in a range of studies (http://gest.umbc.edu/AutoChem/). Code vali-
dation is an important part of this process. The AutoChem system has been exten-
sively validated against a wide variety of data from aircraft, balloons, space shuttle
borne instruments such as ATMOS and CRISTA and satellite based observations.
6. Data Assimilation
The information content metrics and uncertainty characterization will be supplied
by the chemical assimilation system, AutoChem. AutoChem
(http://gest.umbc.edu/AutoChem/) is an automatic code generation sys-
tem, documenter and symbolic differentiator for atmospheric chemical modeling
and data assimilation ?;? . An advantage of assimilation is that it propagates infor-
mation from data-rich regions to data-poor regions. Data assimilation also offers a
mathematical framework to check and quantify the chemical consistency of mul-
tispecies observations with one another and with photochemical theory through
the use of objective skill scores. That is, the analysis can examine both the con-
sistency between different instruments observing the same constituent, and the
photochemical self-consistency between multiconstituent observations and pho-
tochemical theory.
7. Automatic Data Compression
After the raw radiance data observed by a satellite is processed higher level one
and two datasets are generated. These higher-level datasets are usually stored at a
uniform precision, where the stored precision is usually significantly greater than
the certainty with which the level one and two data are known. For example, the
data may be stored with eight significant figures when we are only confident in the
first three or four. If the total data volume is small then this does not have signifi-

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 7
cant cost implication. However, when we are dealing with very high data volumes
this does have a significant cost implication for storage and/or data transfer. For
many years now a variety of data compression techniques have been used that
could be adapted to reduce the amount of space required for data storage and time
for data transmission. The degree of data compression can be chosen to make the
compression non lossy for the accuracy characterized by the assimilation system,
i.e. to three significant figures if that is how well we know the variable instead of
to eight or sixteen significant figures if we do not know the variable to that preci-
sion. If it is found at a later date that reprocessing is required then this can still be
done as the raw radiance data is stored to the full machine precision. Automatic
data compression uses the dynamic data concept in the addition of value added
products without incurring prohibitive space requirements.
8. Machine Learning
The whole approach described depends in large part on the integration of a data
assimilation system. When considering data assimilation of atmospheric chem-
istry, one of the computationally most expensive tasks is the time integration of a
large and stiff set of ordinary differential equations (ODEs). However, very similar
sets of ODEs are solved at adjacent grid points and on successive days, so similar
calculations are repeated many thousands of times. This is the type of applica-
tion that benefits from adaptive, error monitored, machine-learning technology.
Our ODE solver already employs adaptive time stepping with error monitoring,
if this is extended to an adaptive use of machine learning then there are literally
massive potential savings in computational expense. A prototype code has been
developed that we would like to extend here for use within the ODE solver. Early
work seems promising that such an approach would work ?;? . A success in this
area would mean a dramatic reduction in the computational cost of assimilation
and hence of the entire dynamic data retrieval control system.
9. Automatic Analysis and Web Site Creation
To facilitate the analysis and scientific usefulness of the modeling and
assimilation system and the dissemination of the data products the sys-
tem includes an automatic web site generator called CDACentral (for
Chemical Data Assimilation Central). An example is available online at
http://gest.umbc.edu/CDACentral/. CDACentral creates a full cross-
linked web site that presents not only the assimilated analyses, the associated un-
certainties, detailed analysis of the uncertainties, assimilation skill scores, but also
a break down of all the continuity equations and the contribution of each individ-

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
8 D.J. Lary
ual term to the overall continuity equation. It is easy to navigate to a given time
period or constituent by using the site’s javascript navigation bars. This allows
detailed mechanistic studies to be performed. For example, the next subsection
describes how the system has been recently used to show the often unrecognized
role of halogen chemistry in the free troposphere.
9.1. A Case Study: Chlorine Oxidation of Methane in the Free
Troposphere
Atmospheric methane is a key greenhouse gas. Methane and hydrocarbon oxida-
tion are some of the most significant atmospheric chemical processes. The hy-
droxyl radical (OH) is an important cleansing agent of the lower atmosphere, in
particular, it provides the dominant sink for CH4 and HFCs as well as the pol-
lutants NOx , CO and VOCs. Once formed, tropospheric OH reacts with CH4 or
CO within seconds. It is generally accepted that the local abundance of OH is con-
trolled by the local abundances of NOx , CO, VOCs, CH4, O3, and H2O as well
as the intensity of solar UV; and thus it varies greatly with time of day, season,
and geographic location ?.
Methane oxidation is usually initiated by hydrogen abstraction reactions such
as
OH + CH4 ! CH3 + H2O (1)
O(1D) + CH4 ! CH3 + OH (2)
Cl + CH4 ! CH3 + HCl (3)
Br + CH4 ! CH3 + HBr (4)
However, the halogen initiation and catalysis of hydrocarbons is not usually con-
sidered in global chemistry models. This is not due to a lack of kinetic knowledge
but rather an assumption that halogens play a minor role outside of the boundary
layer ?;?;?;?;?;? and stratosphere ?;?;?. Figure ?? (b) shows that in the lower strato-
sphere and even in the free troposphere, halogen-catalyzed, and halogen-initiated,
methane oxidation can be important. Halogen-catalyzed methane oxidation can
play a significant role in the production of HOx (= H + OH + HO2) radicals ? in
just the region where it is usually accepted that nitrogen-catalyzed methane ox-
idation is one of the main sources of ozone ?. Aspects of methane oxidation by
halogens has been previously mentioned by ?;? and the mechanism specifically
described by ?.
Figure ?? (a) shows the fraction of CH3 production due to the reaction of
methane with OH as a height time series at an equivalent PV latitude of 74S, i.e.
in the polar vortex edge region. The analyses was produced using the AutoChem

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 9
chemical data assimilation package and observations of methane, ozone, nitric
acid, and hydrochloric acid from the NASA upper atmosphere research satel-
lite (UARS). The overlaid dashed red line shows the tropopause as diagnosed
by the WMO lapse rate definition, the solid line shows the temperature minimum.
Although the contribution to CH3 production by the reaction of Cl atoms with
methane in the troposphere (below the dashed red line) is usually considered to
be unimportant the analysis produced by data assimilation shows that this is not
true (panel b). Every spring the production of CH3 due to the reaction of Cl with
methane can contribute up to 80% of the total rate of CH3 production. Likewise,
the hydrolysis of BrONO2 alone can contribute more than 35% of the HNO3 pro-
duction rate in the free-troposphere ?. Comprehensive results from the chemical
assimilation are available online at http://gest.umbc.edu/CDACentral/.
Reaction (2) is most significant in the tropical upper-troposphere where it con-
tributes up to 7% to the initiation of methane oxidation for much of the year as
can be seen in the analysis presented in the CDACentral website. Reaction (4)
plays a negligible role and is just included for the sake of completeness.
In this study sulphate aerosol observations from SAGE II ?;?;?;?;?;? and
HALOE ?;?;?;? were used, ozone observations from UARS ? MLS v6 ?;? ,
HALOE v19 ? , POAM, ozone sondes and LIDAR, nitric acid observations from
UARS MLS v6 ?;? , CLAES, ATMOS, CRISTA ? , ILAS ? and MOZAIC aircraft
? , hydrochloric acid observations from UARS HALOE and ATMOS, water obser-
vations from UARS MLS v6, HALOE v19, and ATMOS, methane observations
from UARS HALOE v19, ATMOS and CRISTA were used. All though the bulk
of these observations were in the stratosphere a significant number of satellite ob-
servations were available for the free troposphere down to 5 km, and from sondes
and aircraft data is also available below 5 km.
The major uncertainty in the calculations just presented is the exact chlorine
loading of the free-troposphere. UARS/HALOE did not make a significant number
of measurements in the free-troposphere, and even when it did the altitude resolu-
tion is only 3 km. In the type of objectively optimized earth observations system
envisioned here this type of information can be fed back to the earth observing
system via the SGM to direct further observations to be made, for example by
sub-orbital platforms such as the UAVs (unmanned arial vehicles). It can be seen
that the active synergy between observations and modeling via data assimilation
can facilitate scientific insights. If this synergy is extended to include a dynamic
direction of observations based on objective measures routinely produced by data
assimilation it can be seen how we have a sound strategy for focussing on the key
scientific issues.
The example we have chosen is deliberately a little controversial. The point be-

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
10 D.J. Lary
ing that ‘conventional wisdom’ can make assumptions that do not square against
a large body of observations. The example chosen was from earth observation but
aircraft data and the beautiful ATMOS data set show exactly the same thing. In
addition, one of the purposes of using assimilation is to validate the model, es-
pecially when using high quality in-situ data such as from aircraft. For example,
one can not explain the precise shape of the OH and HO2 diurnal cycles observed
from aircraft in the upper troposphere and lower stratosphere if halogen chemistry
is not used. ATMOS and satellite data also strongly point to the same end. In other
words observations from aircraft, ATMOS, and more than a decade of earth ob-
servation agree with the model used based on well established laboratory kinetics
and disagree with the conventional wisdom that says halogens do not play a role in
the free troposphere. The data speaks strongly against this ‘conventional wisdom’.
10. Conclusion
A schematic overview of the objectively optimized earth observation system envi-
sioned is shown in Figure ??. The elements of the dynamic data retrieval control
system can help in objectively planning mission goals, in the cost effective oper-
ation of future optimized earth observing systems, and for scientific analysis and
dissemination. During the planning stage the objective measures of information
content are invaluable in determining what the instrument capabilities should be.
During the operation of future earth observing systems the dynamic data retrieval
control system could dynamically adapt what measurements are made, where they
are made, and when they are made, in an online fashion to maximize the informa-
tion content, minimize the uncertainty in characterizing the systems state vector,
and minimize both the required storage and data processing time.
The same technology could be applied to the analyses and design of ground
based pollution monitoring networks to provide regular pollution analyses. These
could then be used for epidemiological studies in the precise quantification on the
impacts of pollution on human health. For example, it was noted by Shallcross
(personal communication) that high levels of benzene were associated with high
hospital admissions of cardiovascular conditions.
At a more basic level the idea of symbiotic communication and dynamic data
could be used in many applications to optimize monitoring and observing systems.
The ideas of automatic code generation and automatic documentation to facilitate
system implementation on a variety of hardware is also of quite general applica-
bility. As is the concept of automatic data compression to minimize the required
cost of both storage and dissemination.
Higher level goals could also be specified such as the remote identification of

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 11
sites of likely malaria outbreaks. By facilitating the early identification of poten-
tial breeding sites of major vector species before a disease outbreak occurs and
identifying the locations for larvicide and insecticide applications. This would re-
duce costs, lessen the chance of developing pesticide resistance, and minimize the
damage to the environment.
Acknowledgements
It is a pleasure to acknowledge: NASA for a distinguished Goddard Fellowship in Earth
Science and for research support; The Royal Society for a Royal Society University Re-
search Fellowship; The government of Israel for an Alon Fellowship; NASA, NERC, EU,
and ESA for research support.

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
12 D.J. Lary
Fig. 1. It would be very useful for an earth observing system to dynamically track evolving features.
For example, on the left the sharp gradient in NO (nitric-oxide) at the terminator can be seen. On the
right a visualization of both the tropopause and the polar vortex can be seen, both are important mixing
barriers.
Fig. 2. A schematic of the NASA Earth Science Research Satellites Currently in Orbit.

March 13, 2008 10:54 WSPC/Trim Size: 9in x 6in for Review Volume rs2004rv
Data Assimilation and Objectively Optimized Earth Observation 13
Fig. 3. Schematic overview.
(a)
1992 1993 1994 1995 1996 1997 1998250
300350
400500
600700
8001000
12001400
160018002000
22002400
Year
Potent
ial Tem
peratu
re
% of prod for CH3 due to OH+CH4→H2O+CH3 Bimolecular
1020
3040
5060
7080
90
(b)
1992 1993 1994 1995 1996 1997 1998250
300350
400500
600700
8001000
12001400
160018002000
22002400
Year
Potent
ial Tem
peratu
re
% of prod for CH3 due to Cl+CH4→HCl+CH3 Bimolecular
10
20
30
40
50
60
70
80
Fig. 4. Atmospheric methane is a key greenhouse gas. The main loss of methane occurs through the
reaction of methane with OH to produce CH3. Panel (a) shows the fraction of CH3 production due to
the reaction of methane with OH as a height time series at an equivalent PV latitude of 74S, i.e., in the
polar vortex edge region. The analyses was produced using the AutoChem chemical data assimilation
package and observations of methane, ozone, nitric acid, and hydrochloric acid from the NASA upper
atmosphere research satellite (UARS). The overlaid dashed red line shows the tropopause as diagnosed
by the WMO lapse rate definition, the solid line shows the temperature minimum. Although the con-
tribution to CH3 production by the reaction of Cl atoms with methane in the troposphere (below the
dashed red line) is usually considered to be unimportant the analysis produced by data assimilation
shows that this is not true (panel b). Every spring the production of CH3 due to the reaction of Cl with
methane can contribute up to 80% of the total rate of CH3 production. It can be seen that the active
synergy between observations and modeling via data assimilation can facilitate scientific insights. If
this synergy is extended to include a dynamic direction of observations based on objective measures
routinely produced by data assimilation it can be seen how we have a sound strategy for focussing on
the key scientific issues.