7) Estimate of NH Land+Sea temperatures 1900 - today
For this estimate I will use original NH land temperature data for 1900-60 / Land + Sea Surface temperatures for 1960-81 / Satellite temperature data for 1981 – today.
Today SST is often used when reflecting the Earths temperatures. However since I find SST quite chaotic and random before 1960, I will use the original (published 74-84, see part 2) land data series for temperatures before 1960. This choice is also based on the fact, that practically all sources of temperature data shows a minor dip in temperatures 1940-60 rather similar to the original NH Land temperatures shown in part 2. The dip 1940-60 – appears not controversial.
For 1960-81 I will include SST to Land data as done for the Hadcrut temperature series. In short, The overall Hadcrut temperature is made by adding the “Land-Air” series to SST (ocean water) series and weighting each contribution using area. I have tried several SST data sets from Hadley, and in general the weighting 58% SST and 42% Land seems to be near the fractions used for the NH HadCrut Series:
I then use this Land vs. SST weighting to get the temperature series Land + SST up to 1981. From this point I can then add a data series of satellite temperature data (average of UAH + RSS NH):
The dotted white line is the resulting NH estimate. The stitching in 1981 is based on data from 5 year averaged temperature graphs. This means that the 1981 for both datasets of the stitching represents average of 1979-83 data. It's unlikely that stitching over an averaged five year period would be influenced by random noise.
Lets compare this estimated NH temperature series with hadcrut3v, perhaps the most used temperature series:
For 1920-today, Hadcrut3v NH shows an approximately 0,33 K warmer trend than the NH estimate of this writing. The extra warming of the Hadcrut temperature appears spread out over all years after 1920. (For the ERA-40 data we saw a 0,32K difference to CRU over the shorter interval 1958-2001)
The GISS land + sea has a warming trend 1920-today approximately 0,38 K warmer than the NH temperature estimate of this writing, but no less than 2 thirds of this deviation occurs after around 1980, so it’s a deviation compared to satellite data mostly.
In general, the NH estimate of temperature from this writing shows less warming than the often quoted sources of temperature data:
Even though my starting point with this article was to examine the changes done to temperature data before 1984, it is evident that a significant part of the deviation occurs after 1980, in the years of satellite data.
So I decided to look closer into the years after 1980.
8) The Land-Ocean temperature equilibrium, and UHI.
The satellite NH (Land + Sea) data has smaller warming trend than GISS, CRU, Vinnikov and NCDC land and land+SST series. The smoothed CSST curve resembles the temperature data obtained from satellite (UAH + RSS).
BUT: Satellite data represents both Land temperatures and Ocean temperatures – and yet they resemble conventional SST?
Why is the SST trend similar to the satellite land+ocean temperature trend? Satellite temperatures and SST has one thing in common: They are without the warming error from the land/city temperature stations, UHI:
Could the decadal difference between land temperature trend and sea surface temperature trend to a significant degree be caused by UHI?
And one could ask: Is it in fact not rather odd that the gab between land and sea surface temperatures supposedly just grows and grows? – How can still larger difference in temperature physically exist without collapsing to equilibrium temperatures where SST = Land temp?
Fortunately we have a superb UHI-free dataset to help us out here. We have satellite data for both ocean areas and land areas:
For these UHI free data of highest quality, land temperatures and sea surface temperatures does reach equilibrium now and then (landtemp = SeaSurfacetemp), and therefore the long term land trend is only slightly warmer than the ocean trend.
As we saw earlier, land temperatures changes faster than sea surface temperatures:
And in satellite data we see, that the differences in trend away from equilibrium only exists for a shorter while after a temperature change:
In the UHI-free satellite data, we see that just after few years with no strong external effect on temperatures, the land and ocean temperatures returns to equilibrium and the longer decadal trends are locked together:
The satellite data strongly suggests, that the conventional land temperatures mostly measured from cities should have had the same long term trend as the SST to be correct. Even though we have La Nina periods like the 1999-2002 period of significant cooling and years of volcanic cooling, according to temperature stations, the temperatures from land/cities/airports does not return to sea temperature level according to city-measured land temperatures. Not even close.
When splitting up satellite data in Ocean part and Land part, we find a good match in sea surface temperatures between the satellites and the conventional measured SST:
And when the SST matches the satellite/ocean temperature so splendid, this obviously also supports the usefulness of satellite land data - unless satellites temperatures are only reliable over the oceans…!
Bottom line: Satellite data supports the obvious, that land and ocean data will stick together on the long run due to the permanent drift towards temperature equilibrium between land temperature and ocean surface temperatures.
Perhaps we can say:
The trend difference between CRU land/city temperatures vs. UAH land temperatures amounts to 0,103 K/decade of possibly UHI or siting related faulty warming trend 1980-2007 for land areas. Besides UHI, this exatra heat in land/city/airport temperatures might also originate in warming adjustments.
Just for curiosity, what would the “NH ESTIMATE” from this writing look like if we assumed that UHI in general was around just 0,04K/decade? 0,04K/century applied before the UHI free satellite data began in 1980?
Example - If UHI = 0,04K/decade, corrected up to 1980 for the land contribution:
Obviously, setting UHI to 0,04K/decade is a guess. None the less, it is certainly not a better guess to set UHI = 0,00K/decade for land/city stations, and this also shows how extremely sensitive to UHI the temperature trends are. We therefore have to accept that temperature graphs using land-stations is not a robust method to estimate global temperatures.
9) How much land area is there on the Northern Hemisphere?
[Skip this chaptor - it holds an error. I will later write about the GISS-LAND data problems!!]
(I would like some reader comments on this one: Did I overlook something or is this a severe problem for GISS?)
Besides the UHI or adjustments problems for the NH Land/city/airport temperature data, another problem in data catches the eye.
As earlier mentioned, CRU combines land and SST data approximately using around 40% land 60 % SST which matches real fraction of land and ocean on the NH. But take a look at the GISS data and SST:
In the later years, the combined GISS land + SST series is much closer to their land curve than to the SST curves. This means that GISS are combining land and ocean areas as if the NH land area was markedly bigger than the NH ocean area?
Below I have added 4 graphs: 40%, 50%, 60% and 70% Land area NH. The 40% Land area graph illustrates the graph resulting from using 40% GISS land and 60% CSST – the ratio reflecting the surface of the Earth..
In recent years, GISS combines Land and Ocean temperatures as if the land area was up to 67% of the Northern Hemisphere (!?) - thus one can claim, that GISS land data are not so very different from the original temperature published 1974-84 land data, but what does it help if GISS in the process of combining land + SST adds far too much heat?
Here a closer look at the fraction of land-data needed to end up with the combined land+SST data. I have set temperatures 1960 to zero for all the used series (and all series are 5 yr avg.)
(I stitched HADISST 5 yr avg. with Reynoldsv2 SST 5 yr avg.)
The equation then becomes:
- we have to weight the land temperature graph 66.5% to achieve the resulting GISS total. What happens if we use the real world land/ocean surface rations? Then we get:
Thus, the extra weighting of land appears to induce 0,134 K of extra heat to the GISS NH land+SST graph (when comparing 2007 with 1985).
What if we used another start year than 1960? (Off course, the start year should not matter if GISS always used 40% for the land fraction: Then we would find the 40% Land fraction for any start year.)
But lets try with the basis period 1951-80 in stead (This is used in the GISS Land and GISSS total temperature series):
This slightly increases the land fraction of 2007 to 68%. The temperature differences between trends in recent years are so large, that changes in base period etc have little influence.
I am sure that the algorithm or specific method used by GISS to combine Land temperature and SST explains some of these apparently odd findings. But whatever the “algorithm” used by GISS is, can it be justified that GISS gradually weights the warm NH-Land graph more and more? And ends up with around 67% NH land fraction in 2007 although NH only has 40% land? Maybe, this algorithm or method deserves some attention?
10) Final words
I have touched a long row of subjects trying to get a better understanding of climate related temperature measurements.
For all subjects opinions will be many, so its my hope that this writing will just increase the knowledge of the data and problems of climate temperature measurements.
In short, just a few of the subjects mentioned:
Why do the IPCC show mostly NH temperature of older data series, but somehow, the IPCC manages to avoid showing the bulk of the temperature dive after 1960?
Most tree proxy temperature data used by IPCC originates in the NH land area (Mann, Briffa etc). The NH land temperature data shows a temperature decline not only 1940-1960 but the decline strongly continues after 1960. Is it then correct for IPCC to cut off declining tree ring temperature proxies after 1960?
Howcome increasing SST´s around 1940 are explained with issues connected to the buckets used to collect sea water when the Marine Air temperatures very strongly supports SST around 1940?
Due to the above mentioned data problems for SST 1940-60, the different versions of these SST data are to a very large degree based on adjustments and assumptions rather than actual measured data:
Is it then the scientifically best approach to include SST 1940-60 when describing temperature trends?
Why do Hansen/Lebedeff 87 make the significant colour change between green and yellowish for – 0,5K temperature anomaly and not 0? This seems to hide the temperature decline after 1940.
It seems that natural forcing is strongly involved with the temperature changes:
How come IPCC illustrates the natural forcings so extremely inferior to the human related inflfuences on temperatures?
In the Jones-group paper on ERA-40 vs. CRU differences, the central number they presents can be calculated to:
Difference ERA-40 vs CRU NH: -0,26K , difference globally: -0,26K and then difference SH… : -0,39K. (!!)
This speaks for itself.
Further more, these data suggests that deviations for SH are larger than for NH?
It seems that the temperature dives on approx 0,25K are well documented:
How come the adjusters of temperature data in the “RATPAC” product of NOAA have produced a surface temperature graph which is very different from bot land and sea-surface data? What factors do they use to justify the extreme difference and to almost remove the temperature dives after 1960? Its har to understand how the adjuster team can feel confident about their data when it is so diferent to the actual measurements.
How come temperature products from land and sea shows almost same trend over a decadal scale when we only look at the UHI-free temperature sources?
How come the satellite Land temperature shows significantly colder trend than Land/city/Airport temperatures?
On top of UHI problems in land/city/Airport temperatures (like GISS-land), why do GISS mix Land and Ocean data resulting in a combined Land+Ocean graph much closer to the land graph than the ocean graph?
For 2007 (when comparing trends from 1960) One need to use at least 66,5% Land data even though the NH in real world only has 40% land.
From here I need to go through SH and thus global data – I aim to have this ready in September 2010.
K.R. Frank Lansner
Some useful links and references:
NH land data 1974-84 weighted, 5 yr avg:
Estimate Full NH, 5 yr avg:
Other articles by Frank Lansner:
Ice core data does not suggest a strong warming effect of CO2:
IPCC makes unbelievable error when comparing today’s temperatures with medieval temperatures:
It seems that humans will not be able to raise CO2 levels very much more:
Proxy data does not suggest extraordinary warming today:
Scientific results may have changed after IPCC changed viewpoint:
Temperature decline after 1940 hidden:
Water-gate (no basis for positive feedback theory):
Urban Heat Island – a world tour:
Last changed: 10th December, 2010 at 15:25:04