Click images for more details



Recent comments
Recent posts
Currently discussing

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > Satellite vs surface temperatures

Are surface temperature records unreliable and how do we check?

Surface temperature records can be compared to satellite measurements of tropospheric temperature (approx 14.000ft) where convected heat is expected to cause warming. In fact the rate (trend) of tropospheric warming is expected to be higher than the surface warming trend.

Interestingly, this may not be the case.

Much of what follows was done with the Wood for Trees visualisation tool.

WfT has been around since 2008 and you can find out about how it started at this WUWT post.

Some commenters here are uncomfortable with fitting trends to the data so instead let’s smooth out the noise with a 5 year running mean and look at the general shape of the curves:

GISTEMP, HADCRUT, UAH, RSS five year means.

HADCRUT, GISTEMP, UAH, RSS five year means 1979 - present (full satellite record).

Now sharpen the focus with a 12 month running mean:

HADCRUT, GISTEMP, UAH, RSS annual means 1979 - present.

The curves have the same shape but there is a vertical separation. This goes away when the records are re-baselined* for a more accurate comparison. GISTEMP and HADCRUT use different baselines to UAH and RSS which take their mean from the 32 year period of satellite observations**.

HADCRUT and GISTEMP annual means rebaselined for accurate comparison with UAH and RSS.

The correspondence between the satellite measurements of tropospheric temperature and the surface temperature measurements is clear. This indicates that the surface data are broadly correct.

Here's another rebaselined comparison of UAH and GISTEMP from Climate Charts & Graphs.

Again,the satellite record validates the surface temperature record both in interannual variability and multi-decadal trend.

If UHI or other influences on surface temperature measurements were significant, the effect should be visible over the 32 year period of satellite observations.

*This is the basis of the Wood for Trees temperature index.

This explains how the rebaselineing was done.

Specifics on baselines.

**The slight difference in baseline between UAH and RSS is examined here:

UAH and RSS comparison with common baseline.

Apr 16, 2011 at 10:53 PM | Unregistered CommenterBBD

I have come to the same conclusion, guided by you and other knowledgeable people, but from a background that looked at pictures of thermometer sites and wondered what the hell they were measuring. After reflection anjd experiment, I brelised ntht you were talking sense.

However, the real problems are what forces really drive climate. And, even more importantly, what are the local effects.

It is the penalty of the it a Taleb nsituation mwhjere somebody will get wiped out by a once-in-a-million event whilst others rub along with a lot of small events? That is what we need to know from climate science, i would venture to suggest.

Apr 17, 2011 at 10:19 PM | Unregistered Commenterdiogenes

We are asking a lot from climate science. Hence some of the confusion.

The modelled projections of climate change are viewable here:

Pick a dataset, eg IPCC-DDG 4th Assessment Report, then pick an emissions scenario (A1B is popular), then choose your GCM, eg GISS or Met Office/Hadley (UKMO-Had) Select for anomalies or climatology and you can look at what the models project for precipitation, temperature etc to 2030.

I'm not endorsing the models, nor do I claim they are all hopelessly flawed. This is just a link.

Apr 18, 2011 at 12:09 AM | Unregistered CommenterBBD

I'm sorry, I meant to say that the projections are mapped, so you can get an idea of hypothesised regional changes.

Apr 18, 2011 at 12:12 AM | Unregistered CommenterBBD

Carrying on from the Muller thread, you say;

I think you need to back up the assertion that UHI is capable of significantly (ie measurably) warming the global troposphere.

I'm more interested in whether UHI is capable of significantly contaminating temperature records. Locally, that's a given, ie the UHI effect is easily demonstrable, measurable, and arguably visible in individual surface station records. UHI is also arguably a more significant issue to cities than small amounts of global warming, so as cities get denser in terms of both population and energy, UHI effects may warm those cities faster. So people go and install airconditioning, cooling them, but possibly increasing UHI and energy demands. Or, urban planners look at alternative ways to mitigate against UHI. My local Uni (Reading) does some interesting work on this.

UHI effects are supposed to be removed from the surface station records, which is a human process based on certain assumptions that have been questioned. Some of the adjustments to me look questionable given the surface stations are fairly sparse, and as Anthony Watts' work has shown, often poorly maintained. So I'm less confident about the quality or reliability of those records, especialy over long time intervals.

I'm more confident about the satellite records. They're top-down, all(ish) seeing eyes generating data that may be less prone to human bias. They're prone to drift, either orbital or instrumental which gets corrected or adjusted, but they sense more of the world than the surface stations do.

My curiosity is what they may tell us, because they're measuring different things. The satellites aren't measuring surface temperatures, they're measuring the troposphere temperature. The surface stations measure local temperature at 2m + local altitude. The satellites sense at around 4200m, so not the surface. Between the two measurement altitudes, we have the lapse rate complicating things.

To test climate predictions, like climate sensitivity and tropospheric warming we need both accurate, reliable surface records and tropospheric records to see if there's the predicted 1.2x or more tropospheric warming to prove or disprove global warming. If either the surface or the satellite records are wrong, that won't be possible. If there's no tropospheric amplification, then there's no need to panic.

Back to UHI and whether that can cause real rather than synthetic global warming, I don't know. Only 3-4% of the surface is urban, so despite dense urban areas being a few degrees warmer than rural, is that enough to cause global temperature changes? It's probably enough to alter local climates via changes in temperature, wind, precipitation etc. However, that 3-4% seems enough to potentially bias synthetic global warming, if urban or cropland sites are over represented in current temperature series, as Roger Pielke Snr points out here:

Results show that the GHCNv.2 station locations are biased toward urban and cropland (>50% stations versus 18.4% of the world’s land) and past century reclaimed cropland areas (35% stations versus 3.4% land). However, widely occurring LULC such as open shrubland, bare, snow/ice, and evergreen broadleaf forests are underrepresented (14% stations versus 48.1% land), as well as nonurban areas that have remained uncultivated in the past century (14.2% stations versus 43.2% land).

So UHI could be causing 'global warming' in the temperature series derived from GHCN, and we know how independent surface series really are. I'm hoping Anthony Watts and BEST's publications will clear some of this up. Personally, I'd rather see surface records constructed from rural stations where there has been no or little change in land use. Leave the urban ones to measure UHI changes. Problem seems to be if look at just rural ones, there's less global warming, as well as how to use remote sensing to test for tropospheric amplification at those locations.

Apr 18, 2011 at 2:04 PM | Unregistered CommenterAtomic Hairdryer

Atomic Hairdryer

As you can see above, I've tried to demonstrate the close agreement between satellite tropospheric and surface temperature measurements. There seems to be no evidence of UHI causing a significant divergence in trend between the two.

Likewise, the hypothesised tropospheric amplification is absent. Please do take a look.

The Montandon et al. (2011) paper you reference states that:

Results from the temperature trends over the different landscapes confirm that the temperature trends are different for different LULC and that the GHCNv.2 stations network might be missing on long-term larger positive trends. This opens the possibility that the temperature increases of Earth’s land surface in the last century would be higher than what the GHCNv.2-based GAST analyses report.

So GHCNv2 'might be missing on long-term larger positive trends'. Which as Pielke Sr points out, serves to remind us about the missing tropospheric amplification.

However, it does not address the good agreement in interannual variability and trend between satellite TLT observations and surface temperatures as currently presented.

We need to stick to this central point as it it provides a simple mechanism for testing assertions that there is significant warm bias from UHI in the existing surface temperature records.

Please go back to the top post, have a look at the various graphs I have prepared, and see if you can spot this UHI warm bias, which would of course manifest in higher trends for surface temperatures vs TLT.

It's not there, is it? Ergo, UHI isn't introducing a warm bias into the surface temperature records.

Every time we drift away from the specifics I will repeat this argument because I've got a bit weary of people insisting that so much AGW is actually UHI. As I said, there is absolutely no evidence for UHI measurably warming the global troposphere. None. So if the satellite data are in good agreement with the surface temperature record, we know that:

- The warming is real, not UHI warm bias in the surface measurements (anomalies).

- The surface temperature measurements are not significantly in error (good match with satellite TLT data for interannual variability and 32 year trend).

Apr 18, 2011 at 4:17 PM | Unregistered CommenterBBD

Satellite vs surface temperatures: trends compared.

I know, I know. Trends. But let's be open-minded about this.

The start and end points for this comparison are essentially arbitrary. The series runs from 1979 - present because this is the length of the full satellite record.

What you see is what you get.

GISTEMP, HADCRUT3, UAH and RSS trends from 1979 – present.

Decadal trend (degrees C):



UAH 0.14

RSS 0.16

And here's the re-baselined comparison with trends.

The vertical separation between the various records arises from their different baselines. Correcting for this shows the agreement between them more clearly.

The relation between the trends is unchanged, with RSS and GISTEMP virtually identical and HADCRUT and UAH respectively slightly lower.

The highest trend in tropospheric temperatures comes from the RSS analysis of the satellite data but it only matches GISTEMP. It does not exceed it.

CO2 forcing is expected to raise tropospheric temperatures more rapidly than surface temperatures.

This is not evident from the data, which begs the question: is the hypothesised increase in tropospheric warming incorrect, or is being offset by an atmospheric process not currently accounted for?

Apr 18, 2011 at 5:51 PM | Unregistered CommenterBBD

I think you're showing surface and satellite measurements match, which if true presents pro-AGW believers with a problem because the speculative tropospheric amplification is missing, which also means lower climate sensitivity and no need to panic. And also no or less justification for the social policies people want to impose on us on the back of carbon scaremongering.

I don't believe it answers whether UHI has a significant effect on global warming though, or tropospheric heating. If it does, it'll be included in RSS or UAH because as far as I know, neither adjust for UHI.

UHI though is still a way to potentially fix the problem. People have been saying it creates a warm bias, or the adjustments are incorrect. So you could simply re-adjust the data to remove that bias (real or imaginary), make the surface and satellite records diverge and tropospheric amplification appear. The carbon gravy train rolls onwards. Revised surface record may show slightly less warming but the satellite record remains unchanged. AGW still exists, problem solved. They may even be right.

That would be highly dubious, but then some of the people responsible for the surface record(s) haven't given me any good reason why I should trust them, quite the opposite. If Anthony's station survey had shown all the stations CRN1 and well maintained, there would be more trust and less scepticism. They were not, the people responsible for them seemed very blase about the situation and also seemed to have escaped entirely unpunished. Yet these people were supposedly responsible for maintaining the data we're relying on for some of the biggest and most expensive policy decisions in human history. And they wonder why there's scepticism?

Apr 18, 2011 at 6:40 PM | Unregistered CommenterAtomic Hairdryer

Atomic Hairdryer

I don't believe it answers whether UHI has a significant effect on global warming though, or tropospheric heating. If it does, it'll be included in RSS or UAH because as far as I know, neither adjust for UHI.

Forgive me for repeating myself, but there is no evidence - none at all - that UHI is capable of heating the TLT to a measurable extent.

The point is that the UHI is a surface temperature 'problem'. That's why the comparison with TLT temps is a useful way of isolating the UHI effect on surface temps. That's why I selected this comparison as a quick and dirty way of showing the lack of UHI effect.

I don't understand why the point isn't getting across:

IF UHI were pushing surface temps up, THEN they would trend higher than satallite TLT temps.

BUT surface and satellite TLT temps match well.

SO UHI is not a significant contaminant in the surface temperature records.

AND we can see this WITHOUT getting into labyrinthine arguments about station siting etc.

We both recognise and ponder the absence of the expected tropospheric amplification. Particularly as it might mean that something's missing from our understanding of heat transport through the troposphere and on into the stratosphere. With this in mind, we both wonder about true climate sensitivity to CO2. We wonder about attribution and isolating the CO2 signal from other forcings.

If you have read my many comments here on the insanity of UK energy policy you will know that I am with you on the 'weak justification for dismantling industrial society' argument. For example, see my attempts to engage with ZDB here in Discussion (a couple of threads down from here).

In summary, while I 'believe' in the likelihood of AGW being real, I am far from certain that observational evidence backs the consensus estimate for climate sensitivity to CO2 as ~+3C per doubling of the notional pre-industrial level.

Apr 18, 2011 at 7:23 PM | Unregistered CommenterBBD

I still don't think it necessarily proves or disproves UHI. They're still measuring different things. One the boundary layer, the other the lower troposphere. Unless there's an adiabatic boundary between the two, heat generated at the surface will warm the troposphere. It'll do it even easier if it's heat from heat exchangers on the roof of a building above the boundary layer. If cities keep building upwards rather than outwards, it'll increase boundary layer mixing and disrupt wind patterns. Locally it'll have an effect, globally I don't know if we waste enough heat to make any significant difference. Likewise whether ever increasing numbers of ever growing windmills will have any significant effect on vertical mixing, and if so, whether they'd be more useful set to blow rather than suck.

I guess an example of UHI locally that could easily exceed any global warming signal is somewhere like London's Docklands with it's combination of high rise development and datacentres. That area now generates a lot more heat than the buildings it replaced and that kind of urban growth is detectable via satellite. To me, there still seems a lot of uncertainty remaining about UHI, land use changes, aerosols etc before we can really blame it all on CO2, or say we can correctly attribute every factor.

Having said that, I don't entirely deny the surface record given there's also been a number of independent reconstructions, ie Zeke, Mosh etc that broadly agree with the official records. Then again, there are differences between those, as Lucia's been pointing out in her comparisons to each other, and to the simulations. It still doesn't look settled to me.

I still think much of the UHI debate could have been avoided though, if only NOAA had done their job properly and maintained their instruments.

Apr 18, 2011 at 9:13 PM | Unregistered CommenterAtomic Hairdryer


Well, I tried my best to explain. From here on in it would be repeats of repeats.

Good thing I'm not a teacher ;-)

Thanks for your comments.

Apr 18, 2011 at 11:01 PM | Unregistered CommenterBBD

The US surface area is 2% of world surface area. The US has one of the densest thermometer proxy networks.

What are we measuring here?

If our perspective (point of view) is multicentennial, or at the least decadal, the kind of accuracy that Muller is pushing, for example, is absolutely irrelevant.

Apr 19, 2011 at 9:52 PM | Unregistered CommenterShub


The US surface area is 2% of world surface area. The US has one of the densest thermometer proxy networks.

What are we measuring here?

The correlation between near-global satellite TLT and pseudo-global surface temperatures. The degree of agreement is impressive and made me re-think my position on UHI.

Tropospheric temperatures are in good agreement with surface temperatures. If surface temperatures were significantly biased by UHI, they would trend higher than satellite observations of the TLT.

There's not enough divergence between the surface and satellite records to make UHI a game-changer.

Apr 19, 2011 at 10:55 PM | Unregistered CommenterBBD

When we quantify 'UHI', we are trying to quantify the magnitude of its cause.

If we would expect UHI to be a 'game changer' - in terms of discerning a significant divergence between the surface and satellite record - we would be expecting a change in the rate of temperature-impacting urbanization.

Do we know that has, or has not happened?

In fact, how can say anything at all about the magnitude and the rate of temperature-impacting urbanization by looking at a corrected record? Secondly, I find it hard to imagine how urbanization would produce rate changes at all (i.e., produce ever-increasing anomalies every successive year).

Apr 20, 2011 at 12:23 AM | Unregistered CommenterShub

I thnk urbanisation can cause localised temperature anomalies as Anthony has shown a few times, ie temperature spikes after aircon installations or other changes at stations like airports. If those aren't correctly adjusted, are they showing 'urbanisation' effects, or global warmiing? Or, if there's a warm bias in temperature records caused by a preponderance of urban or cropland stations, is that biasing temperature records?

If it is, then removing any warm bias may alter the correlation between surface and satellite and support tropospheric amplification, as the pro-AGW side predicts. Personally I'm still not entirely convinced or trust the current 'big three' surface records, ie CRU/GISS/NOAA. All of them have demonstrated a warm bias, none of them have demonstrated good data quaility standards. This is where other initiatives like BEST or the Met Office/WMO new data repository may get politically interesting. CRU's under resourced and has a credibility problem, some of the GISS team may also be a polticial embarrassment. If there's an alternative, they can be thrown under the nearest bus.

Apr 20, 2011 at 12:27 PM | Unregistered CommenterAtomic Hairdryer
Apr 20, 2011 at 7:31 PM | Unregistered CommenterShub


In fact, how can [we] say anything at all about the magnitude and the rate of temperature-impacting urbanization by looking at a corrected record?

By comparing it with satellite TLT temps and looking for evidence of divergence (warm bias) in the surface temp records

There isn't much to go on, so UHI doesn't seem to be artificially elevating surface temps.

The increase in urbanisation over the 32 year satellite record has been significant. The associated UHI should show up in higher <I>trends in the surface temperature records over this period.

It really couldn't be simpler.

Atomic H

Personally I'm still not entirely convinced or trust the current 'big three' surface records, ie CRU/GISS/NOAA. All of them have demonstrated a warm bias

Are we sure - the difference between HadCRUT and GISTEMP seems to be GISTEMP Arctic interpolation plus different baselines (which give GISTEMP 'higher' anomalies). When the anomaly time-series are re-baselined the difference is greatly reduced.

Comparing re-baselined HadCRUT and GISTEMP with RSS/UAH on a common baseline gives good agreement across all four time series (see above, top post).

If there's an alternative, they can be thrown under the nearest bus.

This isn't going to happen. BEST will end up broadly in line with the rest, and the satellite TLT record will validate them all. Hansen isn't going to get fired however much one might wish it.

I thoroughly recommend WfT by the way. A very useful tool for sorting out one's own misconceptions.

Apr 20, 2011 at 7:37 PM | Unregistered CommenterBBD


Let's stick to the core discussion. It keeps things simple and on track.

Apr 20, 2011 at 7:39 PM | Unregistered CommenterBBD

I don't think you got my point/question. If there is a magnitude bias, it will not show up when we go looking for a trend divergence.

Secondly, the article from Ross McKitrick is exactly about the urban heat question.

Apr 20, 2011 at 8:15 PM | Unregistered CommenterShub


I don't think you got my point/question. If there is a magnitude bias, it will not show up when we go looking for a trend divergence.

Lost me there. That's why the comparison is done from a common baseline. So magnitude change is properly reflected in comparable anomalies.

Apr 20, 2011 at 8:27 PM | Unregistered CommenterBBD


Also, not sure it makes sense to argue as you do.

We have growth in urbanisation for 3 decades

We have assumed increase in UHI for three decades

We infer a possible increase in T (gridded surface temperature records) caused by UHI contamination

The expected result would be an increase in the frequency of high temperature anomalies

Surface temperature records (gridded anomalies) should demonstrate an increase in trend over the period relative to satellite TLT records (global average temperature anomaly) which are unaffected by UHI

The 'magnitude bias' you refer to would only be invisible in a trend comparison of anomalies if it never changed, surely?

But isn't the argument that the increasing influence of UHI on surface temperatures is being mistaken for/misrepresented as GHG-forced AGW?

Apr 20, 2011 at 8:39 PM | Unregistered CommenterBBD

Shub, Atomic, others

Your comments are welcome. If I've made mistaken assumptions, I want to find out. That's why I posted this up.

Sorry I'm being a bit brief tonight but I am posting in haste while packing. Nearly done though.

Apr 20, 2011 at 10:04 PM | Unregistered CommenterBBD

re BBD

The increase in urbanisation over the 32 year satellite record has been significant. The associated UHI should show up in higher <I>trends in the surface temperature records over this period.

The way I'm looking at it is if UHI has been significant, then it'll show up in both. If it's not significant, then it may only bias the surface record, especially if that's weighted towards urban and cropland stations. I'm still sceptical. Being cynical, why do we need so many thermometers anyway? Surely a couple of isolated, trusted ones should have enough of the global warming signal, Yamal style.

Are we sure - the difference between HadCRUT and GISTEMP seems to be GISTEMP Arctic interpolation plus different baselines (which give GISTEMP 'higher' anomalies).

That comment was aimed more at the demonstrable warm bias of some of the people involved rather than the data. Then again, Lucia's recent post about outliers has given me more to think about.

(what I really want is a couple of 10km^2 grid of precison, low maintenance sensors with dipping ability for the ocean ones, some reflectors to make it an easy satellite target and an automated radiosonde release system. Oh, and a time machine and portable satellite launcher so we've got 100yrs of quality data. Tunguska was not / will not be my test run)

Hansen isn't going to get fired however much one might wish it.

He may, or his growing arrest record may solve that problem. If politicians view him (or Jones) as a liability or an obstacle to policy implementation, then there'll be a bus waiting for him.

Apr 20, 2011 at 11:40 PM | Unregistered CommenterAtomic Hairdryer


The way I'm looking at it is if UHI has been significant, then it'll show up in both.

If you mean both surface and satellite TLT records, why?

UHI is local surface temperature increase. If smeared, Steig-wise, into the global surface temperature measurements, it might distort the big picture.

But it is an artificial amplification. Common sense tells us that UHI doesn't emit enough energy to warm the entire troposphere sufficient to keep it in agreement with artificially magnified surface temperatures.

If the surface temperature records are biased high, why aren't we seeing a trend divergence between them and satellite TLT data?

Apr 21, 2011 at 12:24 AM | Unregistered CommenterBBD

Why? Well, as I see it, it depends on how significant an effect it could be, and it would only need to affect what is being measured. So surface records, and satellite TLT records.

Whether it's enough energy is what I'm less certain about. Taking the figures from the IEA's 2010 key stats, final energy consumption has gone from 4,676 Mtoe in 1973 to 8,428 Mtoe in 2008. If that conversion of energy to work is inefficient, that's generating a fair amount of waste heat that will end up in the atmosphere. There's also the other anthropogenic effects from land use changes as well. If those are warming surface to TLT, and the surface record is correct, then I wouldn't expect to see them diverge.

Apr 21, 2011 at 8:59 PM | Unregistered CommenterAtomic Hairdryer