Latest News (hidethedecline)
Original Temperatures: Turkey
|Posted by Frank Lansner (frank) on 24th December, 2013|
|Latest News (hidethedecline) >>|
Data Sources for Turkish temperature data:
1) Meteorological yearbooks 1929-39, 41-49, 55,62-69,71
2) WWR Turkey to 1970 (mostly 1951-70)
3) GHCN v2 Raw
4) Tutiempo and ECA web sites
In my previous writing on Turkish temperatures, “RUTI TURKEY”, I worked with GHCN V2 raw temperature data. A sad experience because GHCN´s temperature data is so dramatically limited just to show short periods of data, especially for datasets of rural nature.
From “RUTI TURKEY”: Countless rural Turkish data sets are cut down to the short period 1960-90. Only for larger city stations do we find a few datasets 1930-2012.
Therefore I have been rather excited to see what data it is that GHCN has avoided. In this writing I have therefore used GHCN V2 raw temperature data and extended with raw data from other sources (Met. Year books, WWR and Tutiempo).
Doing so I find that that GHCN V2 raw for turkey does not only have an issue with massive data limitation, it also appears that GHCN raw is not that “raw” or “unadjusted”. In some cases GHCN raw versions are indeed adjusted versions:
GHCN raw data in some cases have been adjusted so that data until 1960 is normally raw data, but the 1960´ies and even more 1970-90 GHCN data are sometimes warm adjusted.
As we can also see above, the ECA&D datasets in many cases for Turkey are often short. It seems that GHCN staff thinks that 1960-90 is the best quality of data from Turkey while ECA&D seems to find mostly the 1991-2012 period useful. However, as is normally the case, the adjusted ECA&D temperature data often show warmer data 1990-2012 than Tutiempo.
Despite the above observations, several datasets from Turkey show little or no difference between GHCN and the 3 sources of original temperature data. Stations like this (with smaller adjustments of data) are typically coastal stations or urban stations that normally have more heat trend than locations in shelter of ocean air. As we see in general, it is mostly stations located in shelter of ocean air that has its data adjusted. And thus, it is the typically cold trended areas that mostly have its data adjusted.
Approach to restore long temperature series for Turkey 1900-2007
For GHCN data isolated, the GHCN adjustments gives more heat trend.
Since Tutiempo data normally do not seem warm adjusted like GHCN (See ORI-TEMPS introduction), then a combined temperature series going from the warm adjusted GHCN data to Tutiempo data will create a false and too cold temperature trend. Therefore we cannot just extend warm adjusted GHCN with Tutiempo data to get longer temperature series.
The easiest approach to get longer temperature series for Turkey is to simply exclude the GHCN data when it appears to be warm adjusted like in the examples above. That is, we could just use Original data, WWR and Tutiempo.
However, these datasets often have some gabs mostly in the 1960´ies and 1970´ies that can be filled out by interpolating the particular years from the GHCN series.
Thus, the procedure of restoring longer Turkish temperature series is:
1) When GHCN data appears not warm adjusted:
Use GHCN as is and extend with original data from meteorological year books + data from WWR + Tutiempo data when possible.
2) When GHCN data appears warm adjusted:
Use Original temperature data from meteorological year books + data from WWR + Tutiempo data.
Gabs will be filled in by interpolating from GHCN as long as the size of GHCN adjustment appears clearly.
The use of GHCN data in 2) to fill gabs generally will not change long term trends but is done simply to achieve more continuous data series. If no gabs where filled by GHCN data, trends would generally remain the same.
Approach to restore long temperature series for Turkey 2008-2012
As mentioned, to a large extend I have used tutiempo data until 2007 for Turkey.
But from 2008, Tutiempo data for Turkey appear scrambled, a bit like “Spanish Scrambling” described in ORI-TEMPS SPAIN.
Notice how Tutiempo data often appear broken from 2007 to 2008:
Not rarely, Tutiempo temperature data seems to explode from 2008 and foreward. As explained in ORI-TEMP Introduction, tutiempo do not receive requests to change data. That is, it should not be possible for Tutiempo data to have longer datasets back in time with adjusted data. However, it certainly appears as if the temperature data tutiempo has received for Turkey via WMO has issues from 2008.
Fortunately, in many cases Tutiempo data do not seem scrambled after 2008 and therefore it is still possible to use some of the Tutiempo datasets fully until 2012. Examples with no obvious problems for recent Tutiempo data:
For these stations, Tutiempo temperature data also after 2008 seems to be confirmed by other data sources. Notice that when Tutiempo data trend 2005-2012 can be confirmed by other data sources, the temperature peak around 2009-10 is much smaller.
Turkish temperature trends.
As described in the ORI-TEMPS Introduction, temperature trends since around 1930 seems to depend on the geographic location for all countries examined around the world, which is also described in the RUTI writings. The most fundamental types of areas are
1) Coastal areas
2) OAA, Non-coastal Ocean Air Affected areas, typically hill sides facing winds from oceans or large flat areas without any shelters against ocean air. (OAA areas are shown in yellow).
3) OAS, Ocean Air Shelters, that is, areas in shelter of the winds originating from oceans or lakes. (OAS areas are shown in blue).
Turkish temperature stations analysed for this writing. Red are Coastal stations, Yellow-black are non-coastal but OAA (Ocen Air Affected), and finally we have the blue stations OAS (Ocean Air Shelter).
OAS areas are numerous in Turkey because of the high altitude areas even close to the coast. This gives many areas of Turkey that are valleys and thus in shelter of ocean air.
Due to the Turkish geography, it is likely that a rather large part of Turkey (the OAS areas) has little or no heat trend after 1930. See ORI-TEMPS THE ALPS for a similar scenario.
Coastal temperature trends. Results suggests temperatures for the recent decade around 0,8 K warmer than the 1940-60 warm period.
The Turkish inland areas affected by ocean air show temperature trends similar to the coastal trend.
Some Turkish Ocean Air Shelter stations (OAS)
Turkish OAS stations shown as best fit to the Aksehir station.
For Turkey we see the classical differences between Coastal areas and then the areas in shelter of ocean air (OAS). Recent decades for the OAS stations had similar temperatures as observed in earlier periods of the 20´th century while coastal stations show a heat trend of around 0,8 K since the middle of the 20´th century.
Also notice that the areas less affected by the ocean air show more variance than the coastal stations. Water do have a buffering effect on temperatures, and therefore this difference is not that surprising.
If we focus on 10 year trends, we see that Coastal stations and Ocean Air Affected areas show similar long term trends. For the Ocean Air Shelter areas we see that recent temperatures are similar to 2 warmer periods earlier in the 20´th century.
Red graph represents the “BEST” version of Turkish national average temperature trend. I have inserted Coastal trends and OAS data series – also 10 year average graphs – for Turkey. BEST national temperature trend for Tuker resembles the Coastal trends and the OAA trends. The more cold trended OAS data seems not to have much impact on BEST data. One has to note, that a large part of Turkey – probably more than 50% - OAS areas, areas surrounded by mountains that thus makes a shelter against the Ocean Air trends.
As seen for other countries analysed using original temperature data, the BEST versions for the countries appear not to use data from the sites not affected by ocean air, the cold trended sites.
Last changed: 24th December, 2013 at 16:19:22Back