This map is an interface to compare adjustments made by NCDC and CRU to about 2800 weather stations. Zoom in by dragging a rectangle over a region, and then click on a station. The resultant plot shows a stack of 4 graphs. The bottom displays the monthly average temperatures from CRUTEM4 (green) V3U (blue) and V3C (red). The next graph hows the monthly ‘anomalies’ relative to 1961-1990, and above that are the averaged annual anomalies. The top graph shows any adjustments made by V3C (red) and CRUTEM4 (green) relative to the ‘raw’ values V3U.
The size of the red dots indicates the time span of the station. Clicking opens a new window. Close it, then zoom out (click anywhere) and select another station. An example graph from Iceland is shown below.
Temperature data for VESTMANNAEYJA. Anomalies are calculated relative to monthy normal values from 1961-1990. The top graph shows the net data correction/homogenisation applied to the raw values (V3U). Note the 2 stage offsets made before 1968.
P.S. there is a redundacy in some of the US stations as NCDC add a last digit to WMO numbers for nearby stations. For the moment you get the last one – sorry. I can fix it later.
The Urban Heat Island(UHI) effect ‘cools’ the past in CRUTEM4 land temperature series. This may seem counter-intuitive but the inclusion of stations in large cities has introduced a long term bias in normalised anomalies.
CRUTEM4 calculated with (red) and without(blue) the 500 fastest warming stations (mostly large cities)
The reason for this bias is that each station gets normalised to the same 1961-1990 period independent of its relative temperature. Even though we know that a large city like Milan is on average 3C warmer than the surrounding area, it makes no difference to the apparent anomaly change. That is because all net warming due to city growth effectively gets normalised out when the seasonal average is subtracted. As a direct result such ‘warm’ cities appear to be far ‘cooler’ than the surrounding areas before 1950. This is just another artifact of using anomalies rather than absolute temperatures.
I analysed all 6520 stations in CRUTEM4 and identified those stations whose average anomaly increase from the period before 1921 to the period 1990-2015 was greater than 1C . There are 538 such stations, the majority of which are in urban areas. These include places like San Diego, Calcutta, Melbourne, Beijing, Shanghai etc. The full list is here. I repeated exactly the same calculation of CRUTEM4 annual anomalies both with and without these stations. Even without these cities the remaining coverage is almost unchanged at 5982 stations. The results are shown above and below.
Detailed look at the change in global temperature anomalies before 1950. Excluding Cities increases 19th century anomalies and reduces net global warming.
The most rapidly changing stations due to post 1950 urbanisation cause a net reduction in 19th century anomalies of ~0.2C
Detail of recent trenbds in land anol,amies with and without the 538 ‘cities’.
The recent trends are very much smaller – only ~0.02C. The reason for this is simply that the fastest ‘energy’ growth in major cities had already occured before 1990. The overall rise in temperature gets normalised out when calculating anomalies. However its effect simply reappears as an excess ‘cooling’ in the 19th and early 20th century.
Past studies (including mine!) have claimed that the UHI effect is very small partly because they focus on recent trends in temperature anomalies. However this is not the case once the pivot effect of the normalisation period is included. Overall I find that the UHI has increased global warming on land by about 0.2C since 1850 by artificially supressing land temperature anomalies pre-1960.
The figure below shows all the major land temperature (anomaly) series overlayed. There are still some slight differences in normalization. Berkeley use individual station baselines, GISS use stations normals in 1951-1980 and CRU 1961-1990. NCDC uses normals relative to the full 20th century shifting values up about 0.15C higher and for that reason is not included below.The V3C offsets are my invention and use station offsets to grid average temperatures with normals in 1961-1990. The core set of station data for all groups are mostly based on those in GHCN V3.
Main temperature series compared. The curves are FFT smoothed fits to each dataset. The grey band is the average of all these individual curves and represents some ‘universal trend’.
We then smooth each dataset based on an 4-year wavelength FFT filter. These are the dashed curves. Finally we average all these together to give one ‘universal’ trend shown as the grey curve. The result is a net 0.8C rise since 1950 over land surfaces
Now we take that Land trend curve ‘L’ and combine it with Ocean temperature data HadSST2 as follows. First, I use the same FFT smooth on HadSST2 to give an Ocean trend ‘O’. Then define a global trend ‘G’ based on the fractions of the earth’s surface covered.
G = 0.69*O + 0.31*L
Now we simply plot G and compare it to the latest annual Hadcrut4 data.
Compare a simple linear combination of annual Land and Ocean temperature trends to HadCRUT4
It is an almost perfect fit. This at least demonstrates an overall consistency between the ocean, land and combined surface data. The different algorithms applied are now all basically in agreement.
What’s left is looking into any systematic biases hidden in the more than 7200 individual station data. These could be caused both for instance by ‘homogenisation’ and the urban heating effect. That is if I can still find the energy to delve into it !