Atmos. Meas. Tech. Discuss.

, 5, C1267–C1285, Atmospheric
2012 Measurement
www.atmos-meas-tech-discuss.net/5/C1267/2012/ Techniques
© Author(s) 2012. This work is distributed under Discussions
the Creative Commons Attribute 3.0 License.

Interactive comment on “Performance of

a low-cost methane sensor for ambient
concentration measurements in preliminary
studies” by W. Eugster and G. W. Kling
W. Eugster and G. W. Kling
[email protected]

Received and published: 13 June 2012

Final response – our replies are given in blue color between the reviewer statements.

C1267

Anonymous Referee #1

General Comments

This manuscript describes the evaluation of a low cost, low performance sensor for
methane to determine the suitability for its use in a survey application. The intention is
to identify some means of surveying extended greenhouse gas emission sources,
such as arctic tundra. A network of low cost sensors could be designed to cover the
areal extent of the source efficiently. Higher performance (and therefore higher cost)
sensors could then be deployed to characterize the most representative points of the
source. The manuscript evaluates the performance of a low cost, solid state, metal
oxide (MO)-based sensor for methane. Two such sensors were deployed for a season
on a floating platform in an Alaskan lake, along with a tunable diode laser (TDL)-based
high performance sensor. The MO sensor output is known to have a dependence on
both ambient temperature and humidity, and cross sensitivities to other potentially
important species, such as CO. The authors have presented their arguments clearly.
The supporting figures and tables are sufficient and the language is clear and concise.
Thanks for this positive assessment of our manuscript.
The paper provides a somewhat rare look at some of the issues inherent in designing
a network to measure extended natural emission sources. This includes the
perspective of determining if less capable and less expensive sensors have any value
in surveying for the best sites to locate more capable sensors. The characterization of
the individual sensor capabilities in the field is quite useful and the analysis of the
initial field data is valuable. In this effort though, there seems to be a need for
definition of the metrics for usefulness. For example, in general terms, how much
performance can be sacrificed and still identify major trends; can diurnal excursions
be measured with what signal-to-noise ratio?

C1268
Reply: In our present text we only examined the mean diurnal cycle over a full
season. It remains an open question for which projects a cheap sensor is suitable
(hence our terminology with “preliminary” studies) and where a full-fledged (optical or
similar) sensor is needed. We will do our best in the revisions to clarify this open
question in terms of the “signal-to-noise” ratio.
The performance of the MO sensors were evaluated with respect to one site and not
generalized. It might also be useful to determine the economic boundary conditions of
the problem for a given site. . . for example. . . how many less-capable sensors can one
afford in order to cover the emission source ? Can the required survey measurements
still be made even with inexpensive sensors?
Reply: We agree that this would be a next step in the application of such a sensor.
The authors may want to include mention of other highly capable but still economical,
laser-based sensors coming available. For example, S. So, A. A. Sani, L. Zhong, F.
Tittel, G. Wysocki, “Laser Spectroscopic Trace-Gas Sensor Networks for Atmospheric
Monitoring Applications”, The 8th ACM/IEEE International Conference on Information
Processing in Sensor Networks, Proc. of ACM, vol. ESSA 2009, (2009) as well as
others.
Reply: We will clarify the term “low-cost” in our revised version (see also our reply to
Reviewer #2) and will also add this reference to the list of other economical sensors.
We were unfortunately not able to obtain an estimate of the costs of the sensor by So,
Wysocki etc. but it is certainly worth mentioning despite the fact that the price tag will
need some extra zeros compared to the solid state sensor.

Specific Comments

There are several points in the manuscript that the authors could provide some
additional clarification:
C1269

*Page 2570, line 24 . . . Cross sensitivities. . . Are there cross sensitivities to any
hydrocarbons or other volatile organics that might also be emitted from the lake or
tundra wetlands?
Reply: We address this question later in the discussion on page 2578 and 2579. We
can add a statement in Material and Methods (Section 2.1) to indicate that we are
going to address this question in our discussion.
*Page 2574, line 24. . . Does the sensor drift in one direction only so that initial and
final calibration measurements will provide a linear interpolation? What is the drift
mechanism?
Reply: The manufacturer does not specify any details on this question. Our
interpretation is the following: the sensor is actively heated, has a large reactive
surface due to the granular micro-crystals of the reactive material (see page 2571,
line 14). Most likely during the use of the sensor there will be some loss of sensitivity
either via the aging of the heated granular micro-crystals or by some minor dust or
reaction product deposition that may reduce the efficiency of the reacting surface of
the sensor. This should increase Rs over time, which translates to an apparent
increasing trend in CH4 concentrations (see Eq. (5)) even if CH4 concentration is kept
constant. We however do not have any separate long-term data of such a sensor kept
at constant CH4 over a longer period, but strongly encourage such additional
measurements.
*Page 2575, line 12 The correlation coefficients obtained for the two MO sensors are
quite low compared to that obtained from the manufacturer’s data plot. Is there any
understanding of why this is the case?
Reply: From the manufacturer’s documents it is not clear, how the relationship of the
sensor on temperature and rH should be explained. Namely, since the sensor is
actively heated one would rather expect that the wind speed or turbulent conditions at
the sensor itself might be an issue since the sensor surface may not heat up to the
C1270
desired temperature under strong wind and high turbulence. At the same time the
heat loss from the sensor is also driven by the difference between ambient
temperature and the unmeasured sensor surface temperature. In our understanding,
the only way to address such issues would be to have the manufacturer produce a
combined sensor where the gas sensor and a temperature and rH sensor measuring
the conditions at the gas sensor’s surface is measured simultanteously. As a first step
in this direction we’ll start a follow-up experiment this summer where we will place a
temperature and rH sensor next to the gas sensor.
*Page 2576, line 17 The authors state they found no consistent lag between the MO
and TDL sensors but reproducible lags seem to have been observed with the diurnal
data trend (Fig 5). A little more discussion here might help.
Reply: We’ll address this issue in the revision, but it is important to take note that we
were unable to clearly relate each single data 30-minute average point from the MO
sensor with the corresponding TDL concentration values (see page 2577, lines 19/20
and our reply to Reviewer #2 on the same issue), but found a realistic mean diurnal
cycle if averaged over the entire measurement period.
*Page 2576, lines 19-22 The magnitude of the observed MO sensor drift is twice that
of the seasonal change. What is the smallest seasonal change that can reliably be
recovered?
Reply: Good question! We argue in Section 3.4 that an initial calibration and a
terminal calibration is required to remove the seasonal trend. So our specification of
the trend given on page 2576, lines 18–22 is the result obtained via this approach
(difference between two values) and hence is not suited to address this question. If
one wanted to use the data in a different way (e.g. not calibrating at the end of the
period and simply correct for the drift values we published here), one would need to
explore the temporal variation of the sensor drift. Currently we do not have the
long-term experience to quantify the minimum trend that could be reliably recovered

C1271

without the calibration suggested in Eq. (6). To be honest, we do not actually

recommend to rely on the assumption that the sensor drift is a constant value and the
same for all individual sensors. One would need to test this before making a qualified
statement to address this question.
*Page 2577, lines 19-20 The R2 values obtained for the concentration data of the MO
sensors with respect to the TDL sensor data were poor. . .supposedly because of
time lags. . ..any idea as to mechanism for the lag? If the lag could somehow be
removed, the R2 values would presumably increase. Could a correlation calculation
be done that would identify a constant time lag that could then be removed?
Reply: With time series we always use the cross-correlation procedure to explore time
lags. In the case presented here the conventional result of synchronized time series
was found with the best correlation at lag 0, hence we did not dwell into this issue in
more detail. Unfortunately, there is not an easy solution to find a better correlation by
using lagged correlation analysis, as can be seen in the figure on the last page of this
document which shows the cross-correlation between MO and TDL concentrations
(each smoothed by a 60-minute running average filter). The red vertical lines are
drawn at 1-day intervals to show the effect of the diurnal cycle, which is also seen in
the cross-correlation plot.
*Page 2579, lines 12-13 CO contamination from forest fires could be well above
normal ambient, depending on the distance of the observation point from the fires
Reply: Yes, we thought the same and hence included lines 5–9 on page 2578 saying
“there is some risk of confounding effects in areas where wildfire or other burning is
present. However, during summer 2011, it appears that the uncertainty of the
behavior of the TGS 2600 at cold temperatures (around freezing and below) led to the
largest discrepancies with the reference concentration measurements”. Since we
could not find any evidence that there was any fire influence at the time where we had
the largest discrepancy between the LGR FMA 100 methane concentration

C1272
measurements and those obtained from the Figaro sensors, we believe that our data
are minimally biased by CO contamination.

Technical Corrections

*Page 2572, line 2 RL should be Rs

Page 2575, line 10 R2 should be Rs
Reply: Thank you very much for pointing out these typos, we will rectify this in our
revised version.
We thank Reviewer #1 for detailed and helpful feeback on our manuscript.

Referee #2

J. Kutzner (Referee)
[email protected]

General Comments

This work reports on the evaluation of a non-specific solid state sensor for the
measurement of methane concentrations in air. Measurements of this kind are
important and have the potential to initiate a new direction in sensor developments; it
would be a breakthrough if low cost solid state sensors could be used to measure the
low atmospheric concentrations with sufficient precision. However, if the issues of gas
species cross-sensitivity, temperature dependence, and relative humidity sensitivity of
these solid state sensors (and others) cannot be solved or if their performance leads
to unacceptable results, it might encourage the development of alternative low-cost
optical sensor technology.
C1273

Reply: We definitely agree

Although the presented research has several shortcomings it might stimulate
discussion about suitable sensor technology for atmospheric research.
Reply: We hope so and will work on the shortcomings in our revision of the
manuscript.
The authors report on atmospheric methane measurements. Methane is an important
greenhouse gas. Unfortunately the authors do not discuss this importance in sufficient
detail and therefore fail to provide the basis for the important contribution that their
work might have for measuring this greenhouse gas. For example, methane could be
compared to carbon dioxide and concentrations versus heat trapping capabilities
could be briefly mentioned. Another important aspect is the short chemical lifetime
(about 12 years) of methane in the atmosphere. Methane might have significant
implications for the Earth’s climate future because rising concentrations have been
reported from different monitoring stations over the last few years. Increased release
from Arctic regions or tropical wetlands could drive a feedback mechanism. Methane
released from hydrates on the ocean floor and the implications of such an event might
also be shortly addressed. These points might be referenced in addition to those
already mentioned by Dr. Peter Werle.
Reply: We understand the critique, and certainly such a general description of the
dynamics of methane could be helpful for some readers. However, we were under the
impression that Atmospheric Measurement Techniques rather focuses on
measurement techniques and less on all aspects why one would want to use these
techniques. We may have been mistaken and will add a short (but not exhaustive)
paragraph on the aspects mentioned by Dr. Kutzner in our revised introduction.
Another important aspect missing in the paper is the clear outline of requirements for
atmospheric methane measurements. Atmospheric methane concentrations are low;
in the 2 parts-per-million range. The challenge for atmospheric methane sensors is to
C1274
measure the concentration accurately but also to measure small changes in methane
concentration, i.e. changes in the 10-parts-per-billion range (or below). This is a
challenge even for high performance optical sensors. The “signal-to-noise
characteristics” of an optical sensor must be in the 1:1000 range to be suited for
atmospheric methane measurements.
Reply: Maybe we should have compared such a sensor with the concept of passive
samplers used for O3 , NH3 , NO2 etc. Although one can sample such a sensor at
relatively high temporal resolution, we may leave a wrong impression by even thinking
of achieving the quality of an optical measurement principle with such a low-cost
sensor. What we will do in our revision is to add a paragraph that more clearly refers
to the concept of passive samplers (where rough longer-term averages, e.g. weeks or
months are targeted, not the short-term fluctuations). Still, we agree that it is a
challenge to measure CH4 with high accuracy.
The authors report decent results that they achieve with “low-cost” solid state sensors
operating in the field. These sensors are not intended to measure extremely small
concentration changes of methane. Therefore the presented study represents a first
approach to evaluate such sensors for such demanding tasks. Only relative
measurements are possible with the solid state sensors and calibration
measurements require a high performance methane analyzer. However, the argument
using many low- cost sensors for large area coverage is attractive. Two solid state
sensors based on the Figaro TGS 2600 were run parallel and the results compared to
a high performance optical methane sensor from Los Gatos Research. Main
problems of the solid state sensors are outlined as cross-sensitivity to other gas
species as well as temperature and relative humidity sensitivity. Cross-sensitivity to
other gas species is discussed in detail. However, the measurement location in
Alaska essentially eliminates cross-sensitivity. Temperature and relative humidity
sensitivity of the sensor seems critical. This is addressed and the data have been
corrected for both. An important aspect here is that the “methane sensor” used for the

C1275

presented work consists of the Figaro TGS 2600 sensor head and also other
components. How was this sensor operated? Important details are missing in the
description of the experimental setup. How is the temperature/humidity sensor
placed? How well does it reflect the temperature of the sensor head? What are the
specifications for the temperature and relative humidity measurements? Are they
relevant at the sensor head location? What are expected uncertainties? The expected
achievable measurement precision based on the determined temperature and relative
humidity sensitivity should be discussed.
Reply: We have similar questions, but some of these cannot be easily answered. The
operation principle is only vaguely described in the documentation of the
manufacturer and it is hence not trivial to claim that the user of the sensor has
obtained full physico-chemical understanding of the working principle (the authors
included); because of the limited detail available from the manufacturer, our work has
a somewhat empirical character. With respect to the location and accuracy of
temperature and relative humidity sensor we come to the same conclusion that maybe
it is more important to measure the temperature and relative humidity next to the
sensor (and hence compromise on absolute quality of measurement for ambient
conditions), but even then it is clear that since the sensor is actively heated, the
temperature of the sensor surface is always different from the temperature of the air
sample. Therefore we wonder why this type of sensor is not constructed with a built-in
temperature and relative humidity sensor in one single housing. Such questions
remain, but we will address the issue in our revisions to clarify to the reader in which
direction the next step of improvements should go. We will also add a sensitivity
analysis (boot-strapping) to quantify how uncertainties in temperature and relative
humidity measurements translate to methane concentrations derived in the way we
suggest in our paper.
The “methane sensor” presented here is the sum of multiple components and, due to
the strong sensitivity of the TGS 2600 to temperature and humidity, the overall

C1276
sensing system might be limited by the ability to measure temperature and relative
humidity with sufficient precision. This is only speculation, but the TGS 2600 might not
have been the limiting component of the sensor.
Reply: With our ANOVA we tried to address this issue and found that relative humidity
explains 21.3% of the variance in the sensor signal, air temperature 12.7%, and
methane 18.0%. So, methane and relative humidity seem to have similar weight in the
signal strength. This suggests that despite all concerns about the strong dependence
of the sensor signal on temperature and relative humidity, there is still a visible
methane signal even at ambient concentrations. We think that it is worthwhile to try to
improve concurrent temperature and relative humidity measurements (we actually
have started with this already in a follow-up project). For the revised paper we will add
a sensitivity analysis (see reply above) to help provide additional insight into this topic.

Specific Comments

This work reports the performance of a solid state based methane sensor during field
measurements. These evaluations are important, however, before a sensor is ready
for these evaluations thorough testing in the laboratory under defined conditions
should be performed.
Reply: We see a shortcoming here that the sensor needs a minimum level of humidity
to provide a signal that correlates with methane concentration (our Fig. 3, basically at
least 35–40% relative humidity). In a previous project we had a similar issue with a
photoacoustic analyser where we actually did such a thorough testing in the
laboratory to only find out that although under laboratory conditions our photoacoustic
sensor was partially able reproduce the CH4 concentration, it completely failed under
real-world conditions. This is a general problem of nonlinear sensors that respond to
multiple influences. We do not believe that a simple linear combination can solve this
issue, hence we have not done this, but we would welcome others to demonstrate that
C1277

this approach does work with such a multi-gas sensor.

Questions of how stable the sensor is and what the limiting parameters are can be
more easily addressed in the laboratory than in the field. For example the question of
sensor pressure dependence is not addressed at all.
Reply: We have ambient pressure data and reran our ANOVA (Table 2) with
atmospheric pressure as an additional factor. In this analysis the explained variance
due to pressure variation is less than 0.04% and hence two orders of magnitude
below the factors that we included in our analysis. We will hence expand Table 2
(where we actually address the question of CO2 ) by also including pressure. Since
this is a negligible component we will only briefly address this topic in the revised text.
What does low-cost mean? As mentioned above, a “complete” sensor should be
discussed and not only a component of this sensor. What would be an estimate for a
methane sensor based on a TGS 2600 element? How does it compare to an optical
sensor, for example the Los Gatos Research FMA 100?
Reply: In the set-up we had, the Figaro sensor is approx. US$ 50 and the
temperature/humidity sensor roughly US$ 700. We are currently experimenting with a
low-cost temperature/humidity sensor that also only costs around US$ 50 but could
be placed next to the Figaro. For data acquisition we used a Campbell CR1000 data
logger, which is currently the most expensive component (order of US$ 1000), but we
also have started to experiment with Arduino systems (http://www.arduino.cc/) that
should allow us to build a system that costs on the order of US$ 200 (without labor).
We realize that we get quite some feedback on the term “low-cost”. We’ll have to
specify this more precisely in our revisions, since we learned that some of our readers
consider a methane sensor that costs US$ 25,000 as a “low-cost” sensor already.
We’ll clarify this in our revision.
A paragraph where optical sensors are compared to solid state sensors would be
helpful. The current trend is to replace solid state sensors with more reliable optical
C1278
sensors. How well is the FMA from Los Gatos suited for atmospheric methane
measurements? What are the specifications of this device?
Reply: It’s probably one of the best and most reliable CH4 sensors on the market, so
we can easily elaborate on this issue and add additional references in our revised
version. We have published quite a detailed investigation on this sensor in Eugster &
Plüss (2010) (referenced e.g. on page 2572, last line) but in the meantime there are
more studies and white papers that could be referenced.
I recommend avoiding the term ‘cross-sensitivity’ for temperature and relative humidity
and reserve this expression to refer to ‘different gas species’.
Reply: We can modify the terminology accordingly.
Sentence in line 19/20: This sentence should be reformulated. Not just increases in
the atmospheric concentration of methane contribute to global warming; the current
atmospheric concentration itself is also a contributing factor.
Reply: Yes this is correct, the current methane concentrations contribute to trapping
heat near the Earth’s surface, and any increases in methane will increase this
contribution. We will reword this sentence to make it more clear.
2.2 Principle of Operation: The TGS 2600 is mounted in a TO5 package. This
package is well known in the laser community and might be a better description than
used by the authors.
Reply: We’ll change the wording accordingly.
Figure 5: How can the mean diurnal cycle data be so good with such small error bars
so as to fall within the size of the marker? How were these data extracted? The Fig. 4
data suggest much larger errors because the data vary between 1.87 ppm and 1.98
ppm.
Reply: These are standard ERRORS (not standard DEVIATIONS), that’s the

C1279

difference. As stated in the text only the MEAN diurnal cycle (hence the standard
ERRORS in the graph) can be extracted from the data, for individual days the scatter
is too large as you correctly saw in Fig. 4.
Two TGS 2600 sensor heads were run parallel. The agreement shown in Figures 4
and 5 are impressive. Does that result from temperature and RH measurements
being measured with the same device? In that case uncertainties in temperature and
relative humidity determination would be propagated in the same way into the
measurement results. This might indicate that the TGS 2600s are better than they
appear in the presented measurements, i.e. when results are compared with the Los
Gatos instrument. A different set-up where every TGS 2600 has its own temperature
and RH measurement device might lead to completely different results.
Reply: This is an interesting thought, but consider that (a) if two Figaro sensors
provide reproducible results in an inter-comparison, and (b) two independent
temperature and relative humidity sensors produce reproducible results in an
inter-comparison, then (c) also the methane concentrations measured in this way
must end up being comparable (reproducible). Since temperature and relative
humidity sensors are actually quite OK for the application of ambient measurements;
hence we do not believe that one would yield “completely different results”. But it will
be interesting to test and we actually will add an additional T/rH measurement in a
future study to address this question.
The traces in Figure 1 are too thick to reveal any sensitivity for methane in the low
ppm concentration range. A look at Figaro’s web site reveals that sensitivity.
Reply: We’ll make the traces thinner.
The Figures 4 and 5 show data with surprisingly good agreement of the solid state
sensor data with the Los Gatos instrument. On the other hand the authors report less
favorable results for the pairwise agreement of all data points from the TGA 2600
sensor with the Los Gatos instrument. A graph with less favorable data would help
C1280
understand the real performance of the solid state methane sensor.
Reply: On page 2577, lines 19/20 we stated that the R2 is 0.195 and 0.191 for the two
sensors. Of course one could add the random scatterplot with weak correlation, but
we feel that it is clear from the low coefficient of determination stated in the text and
the scatter seen in Fig. 4 that an additional graph would not lead to further
understanding of the differences in the sensors.
Discussion of the temperature and RH dependence of the TGS 2600
From (5) it can be concluded that R0/Rs changes by 0.0288 per ppm CH4. The
reported diurnal cycle difference is in the range of 0.01 ppm. For 0.01 ppm CH4
R0/Rs should change by 0.000288. According to (3) R0/Rs changes by 0.0072 per %
RH. What can the CS215-L12 measure? Although the output resolution is specified
as 0.03% the given short term hysteresis of < 1% RH might indicate that this sensor
cannot measure the relative humidity with better than 1% precision in the long term.
That would imply that the sensor cannot resolve changes smaller than 0.5ppm based
on the implemented measurement capability for the relative humidity. According to (3)
R0/Rs changes by 0.0246 per ◦ C. How precise and reproducible can the temperature
be measured? Resolving 0.01 ppm methane would require determining the
temperature with a precision of 0.01 ◦ C in the long term. What are the specifications
on the long-term temperature precision? The temperature in the sensor head is
probably relevant. How realistic reflects the recorded temperature the temperature of
the sensor head? These simple considerations show that very high precision in
determining the temperature and the relative humidity are required. The question
remains why the solid state methane sensor can measure effects of a 10 ppb
concentration variation? Does the averaging compensate for temperature and
humidity uncertainties?
Reply: We see the viewpoint of this reviewer who has a background in high-quality
instrumentation where each single data point conforms to a high level of accuracy and
precision. We also realize that the statistical concepts must be described in more
C1281

detail, although they are standard: In our Fig. 5 we take the point that if the diurnal
cycle is a recurring feature within one and the same statistical ensemble, then the
replicated measurements of that cycle over many days helps to carve out the shape of
the curve (hence the standard error of the mean of each hour in our Fig. 5). So, we
think that more clearly introducing the reader to the concept of passive samplers as a
proxy for accurate measurement obtained over longer time periods (weeks, months)
would help to avoid such misunderstandings.
In conclusion, the presented results seem to indicate that simple solid state sensors
are capable of monitoring the variation of very low concentrations of methane. Open
questions remain which must be addressed in further detailed studies.
Reply: We thank Dr. Kutzner for this careful assessment of our manuscript. We
completely agree with his concluding statement, and if this is the message that the
reader takes home then we were successful to express the potential and limitations of
such sensor, and how they could beneficially be deployed in combination with more
expensive high-quality sensors in field studies.

Referee #3

H. Adam (Referee)
[email protected]

General comments

A strategy of deploying low cost gas sensors for screening studies of Greenhouse
Gases in remote locations is obviously desirable. It’s impressive that the authors were
able to demonstrate good correlation between the TGS 2600 and the LGR reference
analyzer – even if only for a limited sample of the acquired data set.

C1282
Specific comments

1. The significant amount of data manipulation and resulting corrections seem limited
in their utility to temperature and humidity ranges which exist during only a minority of
the time at the high latitudes of the field location. Perhaps these are the only times of
the year during which CH4 variations are important and require to be monitored; but
this point is not emphasized in the discussion.
Reply: Yes, we agree that we only covered part of the year as it is typical in arctic field
research where the ice-free season or vegetation period is of interest. Because CH4
is produced by biological activity which in turn is a function of temperature, knowing
the summer season methane concentration developments is important. We accept
that this should be emphasized in more detail in the discussion and we will add a
short explanation on the seasonal evolution of CH4 concentrations in the arctic
environment. Of course one idea of using low cost low power consumption sensors in
the future would be the measurements over longer times (including winter, if the
sensor really works under cold conditions) to investigate, whether unexpectedly large
CH4 emissions (which translate to high CH4 concentration peaks) can also occur
during periods where one would normally not perform measurements.
2. The cost-benefit of deploying the specific TGS 2600 gas sensor is neither
discussed nor obvious. The balance of equipment and logistics costs associated with
such screening studies at locations like Toolik Lake is considerable. Are the
incremental savings produced by using the low cost gas sensors, with the associated
data corrections and resulting limited data availability worth the effort compared with
using admittedly more expensive CH4 sensors that would require less data
manipulation and provide greater data availability (including over greater range of
temperature and humidity)? Quantitative analysis of such economics would be
desirable.
Reply: Reviewer #2 expressed a similar critique which in our understanding is related
C1283

to the lack of clarity on what “low-cost” actually means. Please see our reply to this
point of Reviewer #2. We think that we can address this issue by explictly referring to
the passive sampler concept used for other components and specify the costs for
reference.
We thank Dr. Adam for this feedback.

Interactive comment on Atmos. Meas. Tech. Discuss., 5, 2567, 2012.

C1284
 0.4
0.3
CCF

0.2
0.1
0.0

−10000 −5000 0 5000 10000

Lag (minutes)

Fig. 1. Cross-correlation between MO and TDL concentrations (each smoothed by a 60-minute

running average filter). The red vertical lines are drawn at 1-day intervals to show the effect of
the diurnal cycle.

C1285