Local and regional temperature series

The previous blog post, Global warming is accelerating, applied four temperature series with global coverage. They were NASA GISTEMP, NOAA Global, Berkeley BEST and HadCRUT4 kriging.

The next blog post will examine a solar cycle model which, according to the authors of two articles published eight years ago, claims that the temperatures in some northern regions are dependent on the length of the previous solar cycle. The model predicted cold temperatures in solar cycle 24, which ended in November 2019 after having lasted for 11 years. It is time to check if the prediction came true. 

The  blog post that you are reading now, examines the local and regional temperature series that will be applied in the next blog post. It shows graphically the temperature series read from these sources, and it compares them against each other. 

The average length of a solar cycle is 11 years. The temperatures are therefore averaged over 11 years in this blog post before plotting and comparison. The average is shown as a moving average, in which each point on the temperature curve shows the average of the monthly temperatures five and a half years before and after that point.

1   The temperature series

1.1   eKlima

eKlima is a web portal which gives free access to the climate database of the Norwegian Meteorological Institute. The climate database contains data from all present and past weather stations of the Norwegian Meteorological institute, as well as data from other institutions (owners) that are allowed to be distributed.

I downloaded the monthly temperatures and the monthly temperature anomalies. The anomalies are relative to the mean in the reference period from January 1961 till December 1990.

1.2   Rimfrost

Rimfrost is a web portal managed by Sigmund Hov Moen which gives access to a unique collection of climate data. It is running on an Amazone server. As recommended, I downloaded a private Rimfrost version to my PC to access the temperatures. The monthly temperatures are as measured or as reconstructed, so I had to calculate the anomalies before plotting and comparison.

1.3   BEST - Berkeley Earth

Berkeley Earth offers temperature data from weather stations, from locations evenly distributed over the globe's land areas, and from many regions. The regions may be countries or greater regions such as the northern hemisphere land areas. 

The temperature data from the weather stations consists of monthly temperatures. The other temperature data consists of monthly temperature anomalies with associated quality data, just as the global BEST temperature series do. The anomalies are relative to the reference period from January 1951 till December 1980.

The locations, which are evenly distributed over the globe's land areas, are selected by either clicking on a map or by specifying a city. The distance between these locations are approximately 200 kilometers. Berkeley Earth explains: 'The Berkeley Earth averaging method takes temperature observations from a large collection of weather monitoring stations and produces an estimate of the underlying global temperature field across all of the Earth's land areas. Once this temperature field has been generated, it is possible to estimate the temperature evolution of individual locations simply by sampling the field at the location in question.' When we specify a city, we get the temperatures series from the nearest of these evenly distributed locations. I name these series 'BEST near' followed by the city name.

The 'BEST near' temperature series all start before 1850. Now, at the end of 2020, they are updated up to an including June 2020. In order to get a correct comparison between the different temperature sources, I use the temperatures up to and including June 2020 for the other temperature series as well.

1.4   NASA GISS

NASA GISS offers temperature data from many weather stations. The user selects how 'raw' the temperatures shall be. The temperatures that I downloaded are both adjusted and homogenized. I name these series 'GISS' followed by the name of the weather station.

The temperatures are monthly temperatures. I therefore had to calculate the monthly anomalies before plotting and comparing them against the other local temperature series. 

2   Plots and comparisons

All temperatures in this chapter are anomalies relative to the reference period from January 1881 till December 1910. The temperature in this reference period is often referred to as the pre-industrial temperature. The plots consequently show the warming since pre-industrial time.

The 11 years moving averages of the temperatures from the different series are usually rather close to each other. The monthly temperatures differ much more.

2.1   Global and regional temperature series

Figure 1 shows that the land temperatures in both Norway including Svalbard and on the northern hemisphere have increased more than the global temperatures have.

Figure 1.    Eleven years moving average of four temperature series with global coverage (GISTEMP, NOAAGlobalTemp, BEST and HadCrut4Kriging), of  the land areas on the northern hemisphere (BEST NH Land only), and of Norway including Svalbard (BEST Norway).

The four global temperature series agree well with each other, and they are much more stable than the two regional temperature series. The main reason for this is probably that the global temperature series include the sea surface temperatures. 

The previous blog post shows that the global warming is accelerating. That seems to be the case also for the land temperatures in the northern regions. Neither Figure 1 nor the figures in the rest of this blog post support the prediction about falling temperatures in solar cycle 24.

2.2   Stykkisholmur and Iceland

Stykkisholmur is located on the west coast of Iceland. The GISS temperature series for this location starts in January 1880. The 'BEST near' temperature series is for a location 77 kilometers east of Stykkisholmur.

Figure 2: Eleven years moving average of two temperature series for Stykkisholmur and one for Iceland.

The BEST and GISS series mostly agree until the mid-1940s. In the decades thereafter, the BEST temperatures fall more than the GISS temperatures. From the mid-1970s the BEST and the GISS temperatures increase about as much. None of them indicates a cooling in the last solar cycle.

The Root Mean Square (RMS) of the monthly differences between the GISS Stykkisholmur and the BEST near Stykkisholmur temperatures is 0.71ºC. The average of the differences is 0.12ºC.

2.3   Longyearbyen (Svalbard lufthavn)

The eKlima homogenized temperature series starts in September 1898. It lacks many of the temperatures in the reference period from January 1881 till December 1910. I use that reference period anyway, both to be compatible with the other plots and because the anomalies it produces seem to be OK.

Figure 3: Eleven years moving average of  two temperature series for Longyearbyen, Svalbard.

The 'BEST near' temperature series is for a location 30 kilometers south southwest of Longyearbyen. In recent years, the Longyearbyen temperatures in winter have been strongly influenced by the lack of sea ice on the fjord. The BEST location is many kilometers inland, and is therefore not so influenced by this. It may explain why the eKlima temperatures have increased much more than the BEST temperatures since the mid-1970s.

The RMS of the monthly differences between the two temperature series is 1.27ºC. The large RMS value may partly be explained by the large average difference, 0.59ºC, between the two series.

2.4   Vardø

Vardø is located at the coast north east in Norway near Russia. The 'BEST near' location is off the Russian coast 82 kilometers southeast of Vardø. The other 'BEST near' location that I could have used, is a little further away from Vardø in the southwest direction. It gives approximately the same results, so I stick to the nearest location.

Figure 4: Eleven years moving average of four temperature series for Vardø.

The eKlima weather station at Vardø Radio has operated continuously since June 1829. The Rimfrost, GISS and eKlima temperatures are close to each other. The BEST temperatures are a little different from them.

The RMS of the monthly differences between the eKlima and the BEST temperatures is 0.79ºC. The average of the differences is 0.03ºC.

2.5   Dombås

The eKlima homogenized temperature series for Dombås starts in August 1864.

Figure 5: Eleven years moving average of  two temperature series for Dombås.

The 'BEST near' location is 40 kilometers southwest of Dombås. Both that location and Dombås are in valleys far from the sea. I therefore expected that the curves in Figure 5 should be closer to each other than they are. The RMS of the monthly differences between the temperatures is 0.57ºC, and the average is 0.10ºC. 

2.6   Oslo

The weather stations in Oslo have been moved many times since 1850. I therefore downloaded the homogenized eKlima temperature series which has monthly temperatures since January 1850.

Figure 6: Eleven years moving average of  two temperature series for Oslo

The 'BEST near' location is far away from the Oslo fjord, 69 kilometer northwest of the city center. I therefore did not expect that the curves in Figure 6 should be that close to each other. The RMS of the differences between the monthly temperatures is 0.56ºC, and the average is 0.00ºC. 

3   Conclusion

I checked the solar cycle model back in 2012 and 2014. Then I mostly used temperature series from eKlima and Rimfrost for the local series. This time, as you will see in the next blog post, I will use eKlima and 'Best near' for the local series. I will use eKlima homogenized temperature series for locations with weather stations that have been moved around since the measurements started.

eKlima and BEST apply different methods for adjustment and homogenization. For Longyearbyen I will apply both of them to see how robust my conclusion with respect to the solar cycle model is.