Imagine, that GHCN took all USA rural stations and cut down to 1960-90. Then took smaller cities limited to 1950-90 or 1960-2010, and then only the largest citiest had long datasets 1930-2010 or longer? Sounds impossible? Well this is what is done for Turkey, Bon apetite.
Turkey is a GOLDMINE of temperature stations, fantastic. A bit like the USA. Every inch of the Turkish soil is covered by a temperature station.
Fig1. The abundance of rural temperature stations in Turkey is absolutely fantastic. The blue icons indicate stations used by GHCN. The green icons indicate that both Hadcrut and GHCN have data for this station, mostly larger cities.
A quick overview of the rural Turkish stations:
Fig2. 82 of Turkey´s 216 temperature series used by GHCN are rural. But to my disappointment not even one single Turkish rural temperature series are made public or used by GHCN outside the period 1960-90. GHCN thus made this it impossible to reconstruct a rural temperature for Turkey. The rural stations may “count” in the statistic of rural vs. urban stations used by GHCN, but they are essentially just not used at all.
In the following, we gradually increase city size to see how this impacts the availability of data.
The first step is then to examine the smallest urban areas – “Sub-Urban” population 10.000 – 49.000:
Fig3. Suburban population 10.000-49.000: Far most (101 !)are treated like the rural series it seems, not available outside the 1960-90 period. In the above suburban series however, a new expanded period is sometimes available, 1950-90. We are allowed to see 10 more years, and also some series are extended to 2010.
One temperature series is extended to start from 1930.
Fig4. For the larger urban areas 50-99.000 still we see a number of datasets limited to 1960-90, but the intervals 1950-60 is added to several datasets, and likewise the 1990-2010 period.
One more temperature series starts 1930.
Fig5. Urban 100-199.000. For this Urban size, still a few 1960-90 cuts, but now the 1950- and 1930 startpoint is dominating.
Fig6. Urban 200-499.000.
Fig7. Urban 500.000 and up. For the largest cities we find one station almost complete available – central Istanbul/Gozt - and still a few 1960-90 cuts that are suburbs to these large cities. In addition an Ankara graph that appears to begin naturally (?) in 1924.
So, what have we learned so far?
All rural stations are cut down to 1960-90. Smaller cities are typically limited to 1950-90 or 1960-2010, and then only the largest cities has long datasets 1930-2010 or longer.
What happens if we compare data 1960-90 between those data from the (often rural) 1960-90 sets with data from the 1950-90 sets and data from the 1930-90 sets etc. Here goes:
Fig8. All datasets has pretty much the same trend. In fact, the 1930-90 sets has slightly cooler trends than average. This is a little unusual, but by now we should think that rural data, urban data shows same trends?
Fig9. - But the apparent similarity between the different categories of city sizes (length of data) vanishes if we add just the 1950-60 period of compare. So is it a coincidence that exactly in the 30 years of rural data availability, there are no differences in trends for city sizes?
A Turkish temperature reconstruction. Obviously GHCN has made it impossible to show a rural temperature trend for Turkey, but heres the best I can do at this point:
Fig10. The 1950-2010 are best represented by the 1950-2010 sets, the smallest towns.
Therefore I Stitch the 1950-2010 trend with the early years from larger cities. This will obviously far from omit urban influence, but reduce it a little.