138: Evidence against temperature adjustments #3 (Scandinavia)

One of the main aims of this blog has been to investigate the extent to which the various datasets in the global temperature record have been adjusted and to ascertain both the impact of these adjustments and their validity. Most of the blog posts for individual countries or territories have sought to quantify the magnitude of these adjustments by calculating two versions of the mean temperature anomaly (MTA) for each region; one based on its raw unadjusted data and a second using Berkeley Earth adjusted data. Then the two are compared and the difference calculated. This difference is often considerable and often shows that the adjustments have increased the amount of reported warming. But I have also investigated the second issue, that of validity. One way to do this is to compare the MTA for neighbouring regions or different data samples from the same region. 

The rationale is as follows. If there are errors in the data that are sufficient to affect the MTA, then comparing MTAs from different samples from the same region, or samples from adjacent regions that would be expected to be almost identical, could highlight the errors. Of course any difference between MTAs from different regions does not prove that the data is wrong; it may be that the regions aren't as similar as one supposed. But if the data is virtually identical then that does suggest both that the temperature trends for the two samples or regions are behaving the same, and that any data errors in the temperature datasets (which are likely to be numerous) are not significant and so are not in need of correction or adjustment.

In Post 57 I used this approach to compare the temperature trends of neighbouring countries in central Europe (Germany, Czechoslovakia, Austria and Hungary). The results showed that if the MTA for a country was determined using data from more than about fifteen different station records then there was little difference between MTAs for different countries, and thus very little error in the MTA of each country. This is because of a property of statistics called regression towards the mean. This basically states that if any dataset contains errors in its measurements (which most data does), and those errors are random in their size and distribution (which they often are), then the errors will tend to cancel each other when you average the data. Moreover, the more data you average, the greater the cancellation of errors and so the more accurate will be the result. If errors don't cancel, then that is because the errors are systematic not random, so the process also helps to identify these as well.

In Post 67 I repeated this process for temperature data from the USA. In this case instead of comparing data from adjacent regions I compared different samples of one hundred stations from the same region: the entire contiguous United States. The result was the same as in Post 57 with each sample exhibiting an identical temperature trend over time with identical fluctuations in the 5-year moving average of the trend.

In this post I will repeat the country comparison of Post 57 but using the 5-year moving average of the temperature trend data from the four neighbouring Scandinavian countries of Norway, Sweden, Finland and Denmark. These trends were determined in Post 135, Post 136, Post 137 and Post 48 respectively. The results are shown in Fig. 139.1 below.

Fig. 138.1: A comparison of the 5-year average temperature trends since 1700 for Norway, Finland and Denmark compared to that of Sweden. The trends for Finland and Norway are offset by ±3°C for clarity.

In Fig. 139.1 I have compared the trends of Norway, Finland and Denmark with that of Sweden. The reasons for choosing Sweden as the comparator were both geographic and practical. It sits between the other three countries and so is a near neighbour for each (Finland and Denmark are not near neighbours so would not be good comparators). But it also has the most stations of the four countries and so should have the most reliable trend.

The data in Fig. 139.1 clearly shows that the trends for all four countries are very similar after 1900 but diverge as one looks further back in time towards 1800. The reason for this is the reduction in station numbers seen in each country as one moves back in time from 1950 (see Fig. 138.2 below). Given that it seems that somewhere between ten and thirty stations are needed in the MTA average in order for the errors to be minimized, we can see from Fig. 138.2 that this condition is satisfied for all four countries after 1890. That is why the MTAs diverge before 1890 but are very similar after that date.

Fig. 138.2: The number of station records included each month in the averaging for the mean temperature trends in Fig. 138.1.

If we just consider the data after 1850 we see that the agreement between trends for the different countries is remarkably good after 1890 (see Fig. 138.3 below). The agreement between Norway and Sweden, and Finland and Sweden are both particularly good to the point of their three trends being almost identical. There is also excellent agreement between Denmark's trend and that of Sweden after 1980 but less so before. This is probably the result of Denmark not only having much fewer stations than the other three countries, but also having fewer than ten stations before 1975.

Fig. 138.3: A comparison of the 5-year average temperature trends since 1850 for Norway, Finland and Denmark compared to that of Sweden. The trends for Finland and Norway are offset by ±3°C for clarity.

Summary

The data in Fig. 138.3 once again demonstrates the futility of temperature adjustments. The fact that the mean temperature anomalies (MTAs) of Norway, Sweden and Finland agree so well for over 120 years from 1890 onwards without data adjustments indicates that the averaging process alone can eliminate most errors.

The Denmark data also adds weight to the conjecture that between ten and thirty stations are needed in the average in order to eliminate most of the data errors. As the error size decreases with the square root of the sample size, an average of 25 datasets should decrease the error size by 80% (reducing each error to a fifth of its nominal value). 

Comparing the data of these four countries in this way also gives us more confidence in the determining the true nature of the regional temperature trend. All the data after 1900 pretty much agree so we can conclude that temperatures from 1900 to 1980 rose marginally by less than 0.3°C and then jumped by about 1°C in the 1980s. But this jump is still only comparable to the size of the fluctuations in the 5-year average.

From 1850 to 1900 both Denmark and Norway diverge from Sweden slightly but in different directions. But this is based on a comparison of only one or two stations in each case and so is not unexpected.

136: Sweden - temperature trends WARMING mostly after 1980

The climate change seen in Sweden over the last 200 years is similar to that seen in Norway over the same period. This is perhaps not surprising given that they are nearest neighbours. The only major difference is the temperature change before 1980. In the case of Norway this was negligible (see Fig. 135.4 in Post 135) and was less than a tenth of the natural variation in the 5-year average. For Sweden there is a modest but distinct upward warming trend from 1860 to 1980 that amounts to about 0.6°C in total. That said this warming is still only comparable with the natural variation in the 5-year average of the mean temperature anomaly (MTA). Then after 1980 the MTA jumps abruptly by about 1°C, just as it did in Norway. The net result is that the temperature trend of Sweden over the last 150 years is one of the few regional trends that actually resembles the global trends published by NOAA, NASA-GISS, Hadley-CRU and Berkeley Earth, but this may not be quite what it seems.

As Sweden has a larger area and larger population than Norway it is not surprising that it has more weather stations. In total Sweden has 25 long stations with over 1200 months of data before 2014 and another 126 medium stations with over 480 months of data (for a full list of stations see here). Their approximate locations are shown on the map in Fig. 136.1 below. The stations are fairly evenly distributed across most of the country, but there does appear to be a greater concentration in the south and only twelve stations (i.e. 8%) are within the Arctic Circle. It does have two of the longest and earliest temperature records though: Uppsala and Stockholm. Both have over 3000 months of more or less continuous data that extend back to before 1760.

 

Fig. 136.1: The (approximate) locations of the 151 longest weather station records in Sweden. Those stations with a high warming trend between 1911 and 2010 are marked in red while those with a cooling or stable trend are marked in blue. Those denoted with squares are long stations with over 1200 months of data, while diamonds denote medium stations with more than 480 months of data.

 

In order to quantify the changes to the climate of Sweden the temperature anomalies for all stations with over 480 months of data before 2014 were determined and averaged. This was done using the usual method as outlined in Post 47 and involved first calculating the temperature anomaly each month for each station, and then averaging those anomalies to determine the mean temperature anomaly (MTA) for the country. This MTA is shown as a time series in Fig. 136.2 and clearly shows that temperatures were rising slowly up until 1980. Then at some point in the 1980s (probably in 1988) the mean temperature appears to increase abruptly by about 1°C.

 

Fig. 136.2: The mean temperature change for Sweden since 1880 relative to the 1971-2000 monthly averages. The best fit is applied to the monthly mean data from 1881 to 1980 and has a positive gradient of +0.46 ± 0.19 °C per century.

 

The process of determining the MTA in Fig. 136.2 involved first determining the monthly reference temperatures (MRTs) for each station using a common reference period, in this case from 1971 to 2000, and then subtracting the MRTs from the raw temperature data to deliver the anomalies. If a station had at least twelve valid temperatures per month within the MRT interval then its anomalies were included in the calculation of the mean temperature anomaly (MTA). The total number of stations included in the MTA in Fig. 136.2 each month is indicated in Fig. 136.3 below. The peak in the frequency between 1960 and 2010 suggests that the 1971-2000 interval was indeed the most appropriate to use for the MRTs.

 

Fig. 136.3: The number of station records included each month in the mean temperature anomaly (MTA) trend for Sweden in Fig. 136.2.

 

The data in Fig. 136.3 indicates that the period of greatest coverage of the country for temperature data is after 1960 with up to 150 long and medium stations in operation at any one time. This drops to about 25 between 1890 and 1960 and to four or less before 1860. This means that the MTA for Sweden before 1860 will be less reliable than its values after 1890. Note that a reliable MTA generally needs data from at least sixteen stations (see Post 57 for evidence). 

Nevertheless, we can calculate an MTA for Sweden back to 1720 and if we do so we obtain the trend shown in Fig. 136.4 below. This appears to show that the MTA exhibited a slow and gentle warming for much of the two hundred years before 1980 before the abrupt jump around 1988, but it can be interpreted differently.

 

Fig. 136.4: The mean temperature change for Sweden since 1720 relative to the 1971-2000 monthly averages. The best fit is applied to the monthly mean data from 1731 to 1980 and has a slight positive gradient of +0.30 ± 0.05 °C per century.

 

If the best fit is performed for the interval 1761-1860 then the gradient is -0.06 ± 0.22°C while the best fit to data from 1861 to 2010 yields a strong positive gradient of +0.99 ± 0.11°C (see Fig. 136.5 below). So changing the fitting interval can change the result; but remember, as Fig. 136.3 shows, the MTA data before 1860 is based on only four station records. Similarly, changing the fitting interval to 1861-2010 also changes the interpretation of the data after 1980. The abrupt jump around 1988 now looks more like a temporary plateauing of the MTA data between 1940 and 1980. So which is correct? The problem is it is difficult to know due to the large natural fluctuations in the MTA data: both could be true or neither could be true. What we do know is that the 1988 jump is seen in the MTA for other countries and for Europe as a whole (see Post 44).

 

Fig. 136.5: The mean temperature change for Sweden since 1820 relative to the 1971-2000 monthly averages. The best fit is applied to the monthly mean data from 1861 to 2010 and has a positive gradient of +0.99 ± 0.11 °C per century.

 

If we next consider the change in temperature based on Berkeley Earth (BE) adjusted data we get the MTA data in Fig. 136.6 below. This again was determined by averaging each monthly anomaly from the 151 longest stations and it suggests that the climate was fairly stable before 1870 but then warmed by over 1.5°C thereafter. In fact the 10-year average suggests a warming of almost 2°C. Moreover the data in Fig. 136.6 is in good agreement with the MTA based on unadjusted data in Fig. 136.4.

 

Fig. 136.6: Temperature trends for Sweden based on Berkeley Earth adjusted data. The best fit linear trend line (in red) is for the period 1876-2010 and has a positive gradient of +1.00 ± 0.06°C/century.

 

Comparing the curves in Fig. 136.6 with the published Berkeley Earth (BE) version for Sweden in Fig. 136.7 below we see that there is good agreement between the two sets of data. This indicates that the simple averaging of anomalies used to generate the BE MTA in Fig. 136.6 using adjusted data is as effective and accurate as the more complex gridding method used by Berkeley Earth in Fig. 136.7. In which case simple averaging should be just as effective and accurate in generating the MTA using raw unadjusted data in Fig. 136.4. In other words, any discrepancy between the adjusted data in Fig. 136.6 and the unadjusted data in Fig. 136.4 cannot be due to the averaging process.

Fig. 136.7: The temperature trend for Sweden since 1750 according to Berkeley Earth.

Most of the (slight) differences between the MTA in Fig. 136.4 and the BE versions using adjusted data in Fig. 136.7 are instead mainly due to the data processing procedures used by Berkeley Earth. These include homogenization, gridding, Kriging and most significantly breakpoint adjustments. These can lead to changes to the original temperature data, the magnitude of these adjustments being the difference in the MTA values seen in Fig. 136.4 and Fig. 136.6.

The magnitudes of these adjustments are shown graphically in Fig. 136.8 below. The blue curve is the difference in MTA values between adjusted (Fig. 136.6) and unadjusted data (Fig. 136.4), while the orange curve is the contribution to those adjustments arising solely from breakpoint adjustments. The change in the overall adjustment from 1876 to 2010 is small, less than +0.06°C while the much larger 0.4°C offset is due to the difference in MRT intervals used to determine the anomalies in Fig. 136.4 and Fig. 136.6 and can be ignored. What this shows is that the BE adjustments in this case are small which is why the data in Fig. 136.4 and Fig. 136.6 agree so well.

Fig. 136.8: The contribution of Berkeley Earth (BE) adjustments to the anomaly data in Fig. 136.6 after smoothing with a 12-month moving average. The blue curve represents the total BE adjustments including those from homogenization. The linear best fit (red line) to these adjustments for the period 1876-1980 has a positive gradient of +0.046 ± 0.002 °C per century. The orange curve shows the contribution just from breakpoint adjustments.

Summary

According to the raw unadjusted temperature data, the climate of Sweden has warmed gradually for over 150 years by about 1.5°C in total (see Fig. 136.4). There is some evidence of an abrupt jump of 1°C around 1988 but this depends on how one interprets the data and the fitting process. The temperature before 1860 appear to have been stable for over one hundred years.

In this instance the adjusted temperature data from Berkeley Earth closely follows that of the unadjusted data (see Fig. 136.6).


Acronyms

BE = Berkeley Earth.

MRT = monthly reference temperature (see Post 47).

MTA = mean temperature anomaly.

List of all stations in Sweden with links to their raw data files.