NOAA supplies two global surface temperature series

NOAA supplies two series with the global surface temperature anomalies. The series use different reference periods. The monthly anomalies in the two series differ a little from month to month also after having been converted to the same reference period, but the mean of the differences is close to zero. I examined this topic in order to decide which of the two series to use, and my conclusion is that both can be used because they result in the same temperature trend. But I am curious as to why NOAA supplies two different series with the global surface temperature.

For many years I have downloaded the NOAA temperature from this NOAA site. After entering the site I tick 'Anomalies and Index Data'. Then I download the temperatures with the default settings, which are Monthly timescale, Global region, Land and Ocean surface, and CSV format. The temperature anomalies are relative to the reference period 1901 to 2000, inclusive. The anomalies had two decimal places up till January 2016, and four decimal places in March 2016. I call these temperatures the NOAA_LandOcean temperatures.

In February I read a blog post which compared surface and satellite temperature series. It used the MLOST record for the NOAA temperatures, and I checked if these temperatures are the same as the ones I used to download. After entering the MLOST site I first tick 'NOAA Merged Land Ocean Global Surface Temperature Analysis (NOAAGlobalTemp)' in the left frame and then 'Ascii Time Series' in the main frame. Then I tick 'timeseries\', and, according to the readme file in that folder, the global surface temperatures are stored in the file 'aravg.mon.land_ocean.90S.90N.{version}.{year_month}.asc'. The temperature anomalies are relative to the reference period 1971 to 2000, inclusive, and they have six decimal places. I call these temperatures the NOAA_Global temperatures.

I first compared the NOAA_LandOcean and the NOAA_Global temperatures which I downloaded in February 2016. Both temperatures series contain anomalies till January 2016, inclusive. I first adjusted the reference period of the NOAA_LandOcean anomalies to be the same as for the NOAA_Global anomalies, and I thereafter calculated the difference between the two series. Figure 1 shows the result.

Figure 1: The difference between the NOAA_Global and the NOAA_LandOcean temperatures. The anomalies were downloaded from NOAA in February 2016.The reference period is adjusted to be 1971 till 2000 for both series.

The black line in Figure 1 is the twelve months moving average of the differences. This average is close to zero, and there is no systematic difference between the two series. The linear trend of the differences is zero. The differences in the monthly temperatures look like random noise, but probably they are not random. The NOAA_LandOcean temperatures have only two decimal places, but the monthly differences are often much greater than 0.01, so the explanation of the differences must be something else than rounding errors.

I then compared the two temperature series which I downloaded in late April 2016. They contain anomalies till March 2016, inclusive. Figure 2 shows the result.

Figure 2: The difference between the NOAA_Global and the NOAA_LandOcean temperatures. The anomalies were downloaded from NOAA in April 2016.The reference period is adjusted to be 1971 till 2000 for both series.

The black line in Figure 2 is the twelve months moving average of the differences. This average is zero. The differences in the monthly temperatures look a little different from the differences in Figure 1, probably because the NOAA_LandOcean temperatures in the March file have much better resolution than they had in the January file.

There is no systematic difference between the two NOAA temperature series, so they are probably based on the same raw measurements and on the same data processing. I am curious as to why NOAA supplies two different series and why they differ as shown in Figure 2.