The NOAA-N Polar Orbiting Weather Satellite
There have been some questions posed recently about how the satellite global temperature record is created by UAH and RSS from the MSU (Microwave Sounder Unit) on weather satellites. For example, Andrew recently posted this comment:
“I keep hearing rumors that the satellite data was “adjusted” to match the surface data better. Is that true? If so, that would be disturbing.”
It is always best to ask the source, so I put the question to Dr. John Christy, who is the lead scientist at UAH that produces this dataset:
“I’ve had some queries on my blog recently that are suggesting that the UAH and RSS satellite data is somehow “tuned” to the surface data, or that the surface data is used to provide some offset function. Given that the MSU looks at microwave emissions from oxygen, essentially a first principles measurement, I don’t see any reason that surface data would be used in any way to adjust the MSU data.
But I figured I’d ask the source, if you’d care to elaborate. If not, no worries.”
To which Dr. Christy graciously responded within a couple of hours:
“No other data are used in the construction. That is why we can do comparison studies without any interdependence.”
So from Dr. Christy’s response it is clear that there is no data sharing or comparative adjustment of any kind between that satellite global temperature record and the land-ocean global temperature record such as is produced by HadCRUT and GISS.
The Advanced Microwave Sounder Unit on NOAA-N satellites
For further reading there is an overview on satellite temperature measurements on and details on the microwave sounder unit both Wikipedia.
UPDATE: An additional question has been brought up related to why the data seems to be missing from the poles. Dr. Christy also responded:
As the spacecraft rolls over the pole it does so at an inclined orbit so
that the highest nadir latitude is about 82 deg with the scanner looking
out a bit closer to the pole. Since we apply the scan line data mostly to
the nadir area directly below the satellite, the actual data only go to
about 83 deg. In the gridded data I interpolate over the pole, but I
wouldn’t trust the data too much beyond 85 deg.
Good to know. I heaved a sigh of relief after reading this.
Thanks, Watts and Christy for everything you have done. Forgive me if I fail to sigh in relief.
Thanks for looking into this.
I, too, have heard recently about some kind of “error” in the satellite measurements that has recently been “discovered” and now the satellite temperatures are supposed to be in closer agreement with the surface measurement. I hope this is not true because as one can see, the actual surface stations are, in many cases, injecting junk data into the averages.
From the Wikipedia entry
http://en.wikipedia.org/wiki/Satellite_temperature_measurements
Thanks to you and Dr. Christy for his prompt response.
(But still I have a couple of questions about how the conversion from MW to temps is done.)
I though the issue was calibration against the surface temperature and possible subsequent adjustments to that initial calibration. If this is the case (and I understand it is) their processing algorithm will make the ‘adjustment’ and no ‘data’ is involved.
Note, the calibration doesn’t have to be explicit. It could through one or more other parameters, like aerosols in the GCMs.
Anthony,
My only question is what was used to calibrate the msu on initial turn up of the sensors. What were the satellites calibrated to for the initial turn up.
Just curious.
Bill
REPLY: Though I don’t know for certain, my engineering side wagers that some sort of microwave radiation source was used in an anechoic radio noise free chamber, such as this one:
http://cegt201.bradley.edu/rfpage/anechoic/anechoic.shtml
Yes, thank you to Dr. Christy for his answer, but — like Evan — I wish that he had been a bit more expansive in his answer. The satellites are not measuring temperature; they are measuring radiances in various wavelength bands. There needs to be conversion; is there not even one set of simultaneous measurements to convert the proxy into degrees of temperature? Even though I had not anticipated ongoing adjustments, I did anticipate some conversion process at some point in history.
The anechoic chamber referenced by Anthony in the comment above is similar to that used to measure the RF radiation pattern of the unit. Parameters such as antenna gain, sidelobes, cross-polarization and beamwidth are characterized and then the data is used in the overall calculations. These parameters are stable and won’t change much over time.
The critical item that requires constant calibration is the unit’s ability to determine the “noise temperature” of the detected RF radiation. According the the NOAA website, the unit is continuously calibrated by referencing a cold and hot calibration source. The cold source is a view of deep space, typically around 4 Kelvin. The hot source is a highly-stable onboard RF noise source at 290 Kelvin. Interpolation is then used to determine the noise temperature of the received signal based on the two calibration sources. The detected RF noise temperature value correlates to the atmospheric temperature.
Lots more information here: http://www2.ncdc.noaa.gov/docs/klm/html/c3/sec3-4.htm
I work in the business, but these aren’t our satellites. However, they all pretty much use similar concepts.
First, I want to emphasize that I consider the satellite folks to be the “good guys” in this sordid affair and my curiosity is not intended to be prosecutorial. And it is worth noting that Dr. Christy indicates clearly that there is not a direct back-and-forth adjustment.
Yet one does tend to wonder in what manner the proxy was calibrated. And, if there is not a 1-1 correspondence, how the progression, well, proceeds.
What caught my eye. actually, was a quote from Dr. Christy in an earlier thread to the effect that the conversion was different at differnt latitudes. And I naturally wondered how this was determined. (I also wondered if the progression of MW intensity and temperature rise is a constant or a curve, and how this was determined.)
Microwaves being a proxy, I assume there must have been an initial correspondence. But I wonder if that very correspondence was corrupted by artifacts in the in situ record and progress at the time.
Evan,
I don’t know if this paper will answer your question. It’s available at ieee.org and they probably want $35 for it unless you’re a member.
Microwave Blackbody Calibration of the TIROS N Microwave Sounder Unit
By Iwasaki, R.S.
This paper appears in: Microwave Symposium Digest, MTT-S International
Page(s): 248 – 250, Jun 1978
Volume: 78 Issue:1
I have a copy but I haven’t gotten around to reading it.
The satellite MSU’s measure brightness temperature at ~60 GHz, not noise temperature. The cold source is the Cosmic Microwave Background at 2.725 K. The hot source is a plate maintained at constant temperature with (ideally) constant emissivity. As I stated elsewhere, there is no low frequency cutoff for black body radiation so all solid objects above absolute zero can emit microwave radiation according to the Planck function.
Temperature as a function of altitude in the atmosphere can be determined because there is a strong oxygen absorption band at 60 GHz. The channels of the AMSU measure frequencies on the wing of this band. However, the problem of extracting a temperature profile is ill-posed because of finite sensor bandwidth and noise. See Grant W. Petty, A First Course in Atmospheric Radiation, Second Edition (paperback), Sundog Publishing, Madison, WI, 2006, pp 228-233, e.g.
Determining the temperature over a fairly broad range of pressure and altitude (lower troposphere, middle troposphere, and lower stratosphere, say) is not as difficult, but it’s still not trivial. Given all the misunderstanding of this subject, something like a review article on the subject by an expert in the field, but written for an audience who aren’t experts would be a good idea.
The satellite directly measures temperature against an absolute standard. One potentially tricky bit comes in the averaging to get *global* warming. One suspects a too clever by half averaging, similar to the too clever by half adjustment of urban sites.
The satellite directly measures the temperature of *something*, but that something is not the atmosphere at a particular height. To derive the temperature at a particular height is model dependent, depends on quantities that are not readily observable. The model can be rather too easily adjusted to give whatever results are desired. Perhaps, if we are only interested in the global anomaly, we can dispense with the model, and just look at the temperature of those wavelengths that have decent penetration to the lower atmosphere, telling us that *something* mighty big has warmed, or failed to warm, even if there is some uncertainty as to what the something we are looking at is.
To allay this suspicion, that the calculations are cooked to get a politically acceptable result, we really should have access to the raw data, and the algorithm by which it averaged.
Seems to me that if we simply took the average observed temperature at wavelengths with good penetration, that would be a good measure of global warming or cooling, and would not be vulnerable to suspicion of too clever by half corrections and adjustments.
DeWitt,
Thanks for expanding on how temperature is determined and for clarifying brightness temp vs. noise temp. Although they are related, brightness temp is the correct term. Regarding the cold source temperature, do you have a reference for the 2.725 Kelvin? In my work measuring antenna G/T, we have always ball-parked the big bang residual at 4 K. While it’s probably close enough for what we do, it would be nice to get it right. In the meantime, I’ll google around.
Yes, this is the sort of answer I am casting around for. I don’t understand it all, but I am beginning to assemble it.
From The Wiki page on the Big Bang:
“In 1989, NASA launched the Cosmic Background Explorer satellite (COBE), and the initial findings, released in 1990, were consistent with the Big Bang’s predictions regarding the CMB. COBE found a residual temperature of 2.726 K and in 1992 detected for the first time the fluctuations (anisotropies) in the CMB, at a level of about one part in 105.”
http://en.wikipedia.org/wiki/Big_Bang
Other sources speak of early measurements in the 3 to 4 Kelvin range conducted in the 1960’s. So my info was only 20 years obsolete!
Mr. Watts,
I wanted to e-mail you, but I can’t find your addy here. Sometimes I really crack myself up. Thought this one might crack you up, too:
http://hennessysview.com/2008/03/10/record-cold-winter-requires-even-scarier-more-draconian-response/
Thanks,
Bill
Evan,
Not sure this is what you’re looking for, but in the microwave region the wavelength is large enough to use an approximation (Rayleigh-Jeans approx) for the Plank function. This approximation shows that the Planck radiance is linearly proportional to temperature. Therefore any trends seen in MSU/AMSU brightness temperatures would be equivalent to trends in temperature of the particular layer sensed.
The “conversions” Dr Christy is likely talking about are corrections to the raw data that are necessary to create a continuous time series from different NOAA satellites. Corrections that need to be made are to account for the temperature changes in the “hot source” Dewitt mentions, drift of the satellite causing a change in local equatorial crossing time (sometimes referred to as the diurnal correction), and a few others. The two I mentioned are latitude dependent.
As far as a “layman’s” explanation the CCSP report is written for policy makers and is a good place to start, any Christy, Spencer or Mears paper has explanations of corrections and a new paper (Randall and Herman, 2008) has a review of corrections for both RSS and UAH in it.
CCSP citation:
Darl, T.R., S.J.Hassol, C.D. Miller, and W.L. Murry (Eds) (2006), Temperature Trends in the Lower Atmosphere: Steps for Understanding and Reconciling Differences, Washington DC
Randall, R. M., and B. M. Herman (2008), Using limited time period trends as a means to determine attribution of discrepancies in microwave sounding unit–derived tropospheric temperature time series, J. Geophys. Res., 113, D05105, doi:10.1029/2007JD008864.
Anthony, it may seem odd that a “polar-orbiting” satellite doesn’t actually cross over the poles, but there are good reasons for it. In point of fact, most earth observation satellites are in “near-polar orbits”, to ensure that the equatorial crossing times are sun-synchronous. In essence, the orbital altitude and inclination are chosen so as to ensure that the satellite’s orbit pattern keeps time with the rotation of the earth, so each ascending or descending node’s ECT is at the exact same (solar) time.
Unfortunately, due to drag, the orbits decay over time, and the ECT drifts slightly from year to year. This is what adds most of the complexity to the MSU data analysis.
Anyway, thanks for posting this. It is nice to see some of the misinformation in the comments in the previous thread get cleared up. Unfortunately, the misinformation and conspiracy theories just keep coming:
To allay this suspicion, that the calculations are cooked to get a politically acceptable result, we really should have access to the raw data, and the algorithm by which it averaged.
Seems to me that if we simply took the average observed temperature at wavelengths with good penetration, that would be a good measure of global warming or cooling, and would not be vulnerable to suspicion of too clever by half corrections and adjustments.
You can get access to the original MSU data. That’s how several different groups (not just UAH and RSS) have come up with their own tropospheric temperature trends. The microwave emissions measured by MSU / AMSU come from a fairly broad region of the troposphere; in order to isolate the signal from the lower troposphere, it’s necessary to somehow remove the signal from higher altitudes. UAH and RSS use quite different methods for doing this, the difference in methods may explain much of the (slight) difference in estimated trends, and it’s not entirely clear which method is better. If you think you have a better way of doing this, by all means go ahead.
Talk about the MSU analysis being “cooked to get a politically acceptable result” is irresponsible, and just feeds paranoia.
New Feb 2008 RSS numbers in also show zero anomaly:
ftp://ftp.ssmi.com/msu/monthly_time_series/rss_monthly_msu_amsu_channel_tlt_anomalies_land_and_ocean_v03_1.txt
Here is the ground track for 1 day of NOAA 18 as seen from above the north pole. I used Satellite Toolkit and their database of TLE .
http://www.flickr.com/photos/24537036@N02/2324711796/
I described SSOs over in the previous discussion of …3 of 4 global metrics
Gentle reader, although any reasonable investigator is ill-disposed, on principle, toward ‘conspiracy theories’, I would like to insert a caveat regarding disinformation.
A successful practice is to feed the unwary good, even much-appreciated information inorder to insert something false that leads one astray.
As an example:
http://en.wikipedia.org/wiki/Stefan-Boltzmann_law
Spot the effluent. Now review the foregoing commentary with an eye for this practice.
J, given that the RSS team has no reason not to believe in AGW and their data mostly agrees with UAH, I’d say that its totally unfounded to be so suspicious.
Are diurnal corrections based on GCM simulations of the diurnal cycle?
JM,
The RSS and UAH groups use different methods to determine diurnal corrections. UAH uses techniques from the satellite data itself (Different methods for Mid tropospheric channel vs Lower tropospheric channel). RSS uses hourly output from a climate model control run.
sooner_wx,
Would the disagreement in decadal trends be from those differing methods?
JM,
Randall and Herman (2008) conclude the differences between the trends (shown in the LT channel over land) of UAH and RSS is primarily due to diurnal corrections.
REPLY: By chance would you have a link to that paper? Thanks, Anthony
Anthony,
Here is the info:
Randall, R. M., and B. M. Herman (2008), Using limited time period trends as a means to determine attribution of discrepancies in microwave sounding unit–derived tropospheric temperature time series, J. Geophys. Res., 113, D05105, doi:10.1029/2007JD008864.
http://www.agu.org/pubs/crossref/2008/2007JD008864.shtml
REPLY: Thanks so much!
I know this is now a rather old thread, since there have been so many new ones recently, but these dated posts about the satellite trends being different from the surface trends seem relevant:
http://science.nasa.gov/newhome/headlines/essd06oct97_1.htm
http://science.nasa.gov/headlines/y2000/ast21jul_1m.htm
Patrick:
Those links are from 1997 and 2000. Over the years since then, several errors were discovered in the methods used by Christy et al for processing the satellite data. The MSU temperature trends now match the surface records quite closely (IIRC, RSS’s trend is slightly higher than GISS, while UAH’s is slightly lower).
In the previous thread, I was defending the proposition that the satellite microwave radiometer data were not “tuned” to match the surface temperature trend. I’m glad that Anthony took my advice and posed this question to Dr Christy. Sometimes it seems like the same arguments keep getting recycled over and over again. Here’s hoping that this particular allegation has now been conclusively put to rest.
J.
I thought that the RSS used a climate model for the diurnal correction, so the question is what are the various inputs used for the climate model and how are the radiative properties of the climate model and how sensitive are they to atmospheric CO2 quantities?
Andrew,
The process of obtaining the temperature from the satellite radiance data is an ill posed problem, i.e. one is inverting an integral and this inversion is especially difficult in the presence of clouds. It is my understanding (based on experience) that the temperture retrieval is an iteration process that uses surface temperature readings in the iteration process. Please ask Dr. Christy again if this is the case or not. If he claims it is not the case, I will dig out the mathematics to prove the point.
Jerry
Andrew,
I just read the info on wikipeidia and it states very clearly that the conversion from satellite radiances to temperature is not a simple process and in fact different results are obtained by different methods. I think Dr. Christy needs to be asked about this in a bit more detail.
Jerry