David Whiteman wrote:
I believe there is consensus that the measurements that cannot currently be made within the specifications of GCOS-112 using any technology include water vapor mixing ratio. Because of that, I believe it is important for this group to specify what intermediate measurement specifications might be useful for assessing climate change while work continues on developing sensors that can meet the GCOS-112 specification.
For example, at the workshop Reinhout Boers presented results that showed the anticipated increases in water vapor over the Netherlands during the next century. An international consortium of climate models has been used to look at the same thing. The results seem to indicate that in general the largest percentage increases in water vapor concentration are anticipated to occur at ~ the 200 mb level with smaller increases both lower and higher in the atmosphere. This implies that trend monitoring of water vapor would be most efficient if sensors were capable of quality measurements at the 200 mb level or approximately 12 km agl. What is the required precision/accuracy required to reveal trends at this level? Again Reinhout presented a framework for determining this using the trend detection work of Weatherhead et. al. (1998). I suggest that this could be an approach for developing a useful set of intermediate measurement specifications for GRUAN. There are several technologies that currently supply quality measurements at the 200 mb level and they include CFH, corrected RS92 and Raman lidar (at night). Perhaps a careful study along the lines described above would indicate that one or more of these technologies is currently able to provide measurements useful for monitoring anticipated trends in water vapor. If GRUAN can develop specifications that can currently be met and that are of sufficient quality to reveal anticipated trends in water vapor, I think it will be in a much stronger position for advocating for resources to support its implementation.
GRUAN Communication Platform 
I concur with Dave Whiteman’s comments.
Regards,
I would like to make some comments on the suggestion of “intermediate measurement specifications”
by Dr. David Whiteman, because I am one of the members who prepared the specifications document.
(The specifications document is currently available from the HMEI website http://www.hydrometeoindustry.org/
–> HMEI Newsletters, Dec. 2008 issue, or from the
GRUAN website in the near future; Holger, Franz, Michael, could you put this document PDF at, e.g., Documentation page, or other easy-to-access pages? Or, to do this, do you need approval by WG-ARO?)
First of all, thank you very much for your suggestion.
I think it is one of the main concerns of all whether we really can start the GRUAN measurements with balloon sensors that satisfy the very-high-level specifications.
But, I am afraid we do not need another specifications document; we have the one that should be enough, and we should go only with this (at least until the 2010 CIMO-GCOS intercomparison campaign).
While preparing this document, many of the members told me that we should not change the GCOS-112 table at all even if some of the numbers may be difficult to achieve with the current technologies; we should show the scientific demand (but I have a comment on this later in this email ).
Therefore, we made only very minor revisions to the original GCOS-112 table.
But, we also put some additional notes for each parameter.
For water vapor, we wrote in page 4:
… it is not necessary for a single sensor to cover the entire altitude range of a balloon sounding.
For example, one sensor is tuned for the lower troposphere, one for the upper troposphere, and one for the stratosphere. Clear documentation about the uncertainty as a function of altitude is essential.
(To tell the truth, this may be a substantial change from what was written in the Appendix 1 of GCOS-112 report; it said, “any proposed instrument or set of instruments would need to be able to operate throughout that range (that range = measurement range).”)
I believe these sentences in the specifications document would eliminate your concern.
If, for example, Vaisala RS92 RH sensors are proved to have specifications that satisfy the requirements between 1000 hPa to 200 hPa at the 2010 CIMO-GCOS campaign, then, the measurements with this sensor (with appropriate operation procedures and with all metadata) between 1000 hPa and 200 hPa are considered as the GRUAN measurements.
This is my understanding.
(Please correct me if I am wrong.)
As for the correction schemes (particularly for polymer sensors), I personally feel there are different ways of thinking about the question, “Can polymer sensors be a reference?”
Some of the past works (including mine) estimated correction schemes by comparing polymer sensors with chilled-mirror hygrometers; from this point of view, chilled-mirror sensors would be the reference with which we monitor the bias and other errors of polymer sensors.
But, I knew from some personal conversations at Oklahoma that there are different opinions.
It is necessary to prepare a clear answer by the 2010 CIMO-GCOS campaign.
By the way, while preparing the specifications document, I felt the need of having a list of reference, peer-reviewed papers that explain each number in the Requirement Table from the scientific point of view.
The GCOS-112 report has no justification on each number, citing no peer-reviewed paper; Its section 3.3 says, “The values in the tables were initially developed by Boulder workshop participants, and subsequently refined by the combined efforts of an invited expert group and solicitation of a broad international group of climate scientists. At the Seattle workshop, these requirements were further refined.”
I did not attend Boulder and Seattle workshops.
So, to tell the truth, I myself cannot explain why these numbers are necessary and enough; all I can say is that these numbers are shown in the GCOS-112 report, and we should not change them without a good reason.
At the next opportunity to revise the table, perhaps after the 2010 CIMO-GCOS campaign, I think we should do this work.
The reference instrument specification document is available here.
I can only imagine all the effort and consideration that went into generating the measurement specifications contained in GCOS-112 and the revisions on the HMEI website that Dr. Fujiwara refers to below. I realize that generating a set of “interim measurement specifications”
would require a careful additional effort that could take energy away from the implementation of GRUAN but it seems to me both a worthwhile and necessary task. Let me comment more specifically on Dr. Fujiwara’s careful response.
It would help greatly if there were citations demonstrating the need for each of the specifications listed in the GCOS-112 table. I support Dr.
Fujiwara’s original desire to have the table elements accompanied by references justifying each one. Specifically, what is the reference that demonstrates a need for 2% absolute accuracy in water vapor measurements from the surface to the LS?
It is realistic to imagine that a reference sensor might consist of a combination of technologies that provide their best measurements at differing altitudes in the atmosphere. This modification of the GCOS-112 language seems very thoughtful. But I don’t believe that the RS-92 radiosonde is able to meet the requirement of 2% absolute accuracy from 1000 – 200 mb. The most recent work in this area that I am familiar with is from a soon to be published paper by Miloshevich et. al. (available on request). This paper concludes that the absolute accuracy of a single
RS-92 radiosonde, after careful processing and application of empirical corrections, is approximately 5% plus a fixed 0.5% RH uncertainty. This fixed uncertainty becomes quite significant at low RH. The application of the time-lag and empirical corrections inherent in the “careful processing” just referred to adjust the RS-92 water vapor measurements by approximately 5% in the lower troposphere and as much as 20% in the upper troposphere. Based on this work, therefore, the RS-92, even after correction, is not a candidate to be a component of a reference radiosonde.
Also the published uncertainty estimates of the CFH range from 4 to 9% in the troposphere and are ~10% in the lower stratosphere. So this technology is also not currently a candidate for the reference sensor.
Does anyone have an idea of a technology that might be able to measure water vapor with an accuracy of 2% in the real atmosphere?
Based on these points, I support both the consideration of useful interim measurement specifications, achievable using currently available technology, and radiosonde correction techniques (including the Miloshevich and Leiterer methods and others as available) as a part of the 2010 CIMO-GCOS campaign.
Thank you.
Intermediate specs – please can we close this discussion
Firstly, my thanks to those who have commented on the report draft. Your comments have been very valuable and are most appreciated.
However, please can we close the side discussion of interim measurement specifications that is being sent to a very large distribution list.
Although scientifically interesting it was not directly discussed at the meeting and therefore is not relevant to immediate activities.
Furthermore, this re-opens closed issues at a time when we need to be moving forwards and not rehashing old ground. As I very briefly outlined in the meeting the choice of a single specification requirement rather than the more usual NWP multi-target approach was VERY DELIBERATE.
Experience of e.g. the AMSU replacement shows that manufacturers will tend to aim to hit the easiest specification. The feeling from a large body of relevant climate experts was that by giving threshold, ideal etc. we would actually make it harder to eventually meet the requirements, and not easier, as manufacturers would in poker parlance stick rather than twist once the easiest spec was met. That is why there is only one number and why it may seem with current technology to be unreasonable / unmeasurable.
Throughout it has been made clear that we are fully aware that current off the shelf / research technology may not meet the stated targets, nor do we expect it to otherwise what would the value of the network be? And we did spend a lot of time (the longest session in the meeting) discussing the intermediate measurements strategy by talking about comprehensive traceable error budgets on current technology and use of best practises.
It is rather chicken and egg … without a network there is no market and therefore the technology will not be developed. So, we start with what we have, prove there is a meaningful market (a number of sites requiring better instrumentation) and then the improved instrumentation to meet the stated requirements will (eventually) come. The alternative is to sit around and wait a potentially inordinate length of time with the network mothballed until technology to meet the aims exists. I don’t believe that to be a remotely viable pathway forwards. Manufacturers are not charitable organisations and don’t have limitless R&D budgets … we need to prove the market exists.
It would, indeed, have been nice to have a peer-reviewed robust quantitatively based reference for every number in the Table, but I have never seen such a table. My limited experience is that they are largely based upon expert elicitation about the questions that the data will be expected to answer and the purposes it will need to serve and that the numbers fall out as a result. This was the basis upon which these tables were constructed based upon input from c.50 participants at Boulder, open peer review, and then refinement in Seattle by c.40 experts (substantial overlap). These experts represented a broad cross-section of the community. I am comfortable that the numbers are defensible on this basis and that if met the network would answer most climate critical requirements as a result.
I hope that this can now close this discussion.
These email discussion thread has become stimulating and interesting may be we need to create a blog at the GRUAN page? Sorry Holger – I know this is more work-:)
From discussions in Norman, OK, I understand most of the stations will be launching a similar sonde – RS92 (see attachment). Given this fact, the initial GRUAN data archive will be “mostly” of RS92. Thus, if some of the corrections mentioned do not require instrument modification, then they can be made on all the sondes that are launched from all the sites from the future GRUAN data archive. For example, the Miloshevich corrections can be made on all the data that are archived if the necessary ground information are recorded during launch. Is this not correct? If yes, then the “corrections” become a research and analysis task and I would support it to be undertaken by GRUAN as a sub group to produce this higher level data quality stream.
I also recognize the fact that what ever corrections are made (if it is algorithmic in nature) to a single sonde type need to be applied across the sonde instrument types that participate.
Belay
Although a suggestion has been made on interim measurement specifications, for me the most important part of this discussion is that it gives us a clearer picture on where we stand for those of us wanting to make a contribution to GRUAN from the measurement end of things. This discussion can certainly be incorporated into the report as comments on the meeting report. Suggestions on interim specifications would have to be taken up at a future meeting if that is desired. For those of us that have not been intimately involved in all of the previous specification discussions this is helpful in informing us on where we stand and injects a dose of reality into what can be done at present. I don’t think the discussion of this topic is, or should be closed, however. This discussion will not keep us from moving forward.