Guest Post by Bob Tisdale
This is an update of the post published a week ago here (WattsUpWithThat cross post is here). That earlier post included “raw” unadjusted data based on an outdated version of the NOAA GHCN dataset. This updated post includes annual land surface air temperature data through 2015 and uses the current version of the “raw” GHCN data…supplied by Zeke Hausfather. Thank you, Zeke. Most of the text is the same. I’ve updated the discussions of the trends and the source of the “raw” data.
I’ve also added a curiosity at the end.
# # #
Quick Answer to the Title Question: Over the long term, the answer is yes, the adjustments to land surface temperature data increase the reported global warming rate, and the differences between datasets are noticeable. Over shorter terms, the answer depends on the dataset.
This is the second in a 3-part series of blog posts. We examined the impacts of the adjustments to global sea surface temperature data in the post here. Where the adjustments to sea surface temperature data decreased the reported long-term warming rate, the adjustments to the land surface temperature data increase the long-term trend.
But, as you’ll see, the adjustments to land surface temperatures have different impacts over shorter time periods.
If you’re expecting the adjustments to the land surface temperature data to be something similar to those presented by Steve Goddard at RealScience, you’re going to be disappointed. Steve Goddard often compares older presentations of global land+ocean data to new presentations so that we see the change in data from a decade or two ago to now. Example here from the April 8, 2016 post here. But, in this post, we’re comparing recent “raw” land surface temperature data to the current “adjusted” data, which is another topic entirely.
There are 4 adjusted datasets presented in the post:
Berkeley Earth – This is the recently created near-land surface air temperature data from the team headed by Richard Muller. The Berkeley Earth data are adjusted for numerous biases and they are infilled. Data here.
NASA GISS – This is the land surface air temperature portion of the Land-Ocean Temperature Index (LOTI) from the NASA Goddard Institute of Space Studies (GISS). GISS does not publish their land-only portion of the LOTI data in easy-to-use form (See note below), so I used the ocean-masking feature at the KNMI Climate Explorer to capture the land-only portion of the GISS data. The GISS land surface air temperature data are adjusted for biases and infilled with 1200km smoothing. The 1200km smoothing does not infill continental land masses completely. This is especially true in the early portions of the data, when sampling is poor.
Note: This is NOT the land-only temperature data from GISS (referred to as their Meteorological Station Data, dTs data) where they exclude sea surface temperature data and extend land surface temperature data out over the oceans for 1200km. [End note.]
NOAA NCEI – The land surface air temperature portion of the NOAA global land+ocean surface temperature data is available from the NOAA webpage here. NOAA adjusts the data for biases. I am unsure if the land surface temperature data NOAA supplies on that webpage have been infilled. The reason: The land surface temperature anomaly map here from the NOAA/NCEI Global Temperature and Precipitation Maps webpage does not appear to be infilled, while the corresponding map here of their combined land+ocean data shows infilling.
UK Met Office – The UKMO uses the CRUTEM4 land surface air temperature data for their combined land+ocean data. The CRUTEM4 data are adjusted for biases but they are not infilled. That is, if a land surface grid is without data in a given month, that grid remains blank. Annual CRUTEM4 data are available here, in this format.
As far as I know, NOAA has not presented their unadjusted GHCN land surface air temperature data as a global dataset in easy-to-use form. However, a number of independent researchers have prepared comparison graphs of the unadjusted and adjusted global land surface air temperature data.
The earlier post included “raw” unadjusted data based on an outdated version of the NOAA GHCN dataset. In response to the earlier post, Zeke Hausfather (now a member of the Berkeley Earth team) has graciously updated his monthly unadjusted global GHCN land surface temperature data through March 2016, using the current version of the GHCN data. (Thank you, Zeke.) See Zeke’s comment here on the cross post at WattsUpWithThat of the original post. The link to that current version of the “raw” data is here.
The differences are minute between the current version of the “raw” unadjusted data and the outdated version (data here) used in the earlier post. You’re more than welcome to download them both and compare them.
Base Years: The first 4 series of graphs are referenced to the WMO-preferred base years of 1981-2010.
IMPORTANT NOTE: This post compares land surface temperature data only. As a result, it does not include any additional warming present in the GISS Land-Ocean Temperature data associated with their masking sea surface temperature data in the polar oceans (anywhere sea ice has existed) and replacing that sea surface temperature data with land surface temperature data extended out over the polar oceans.
LONG-TERM TREND COMPARISON
Figure 1 includes the four “adjusted” land surface air temperature anomaly datasets compared to the “raw” GHCN data. The adjustments to the GISS and NCEI data add about 0.02 deg C/decade to the “raw” data for the period of 1880 to 2015, and the adjustments add about 0.025 deg C/decade in the case of the Berkeley Earth data. The adjustments to the UKMO CRUTEM4 data only add about 0.013 deg C/decade to the reported land surface air warming over the long term.
TREND COMPARISONS FOR 1950 TO 2015
For the period of 1950 to 2015, trends are the same for the “raw” GHCN data and for the “adjusted” Berkeley Earth and UKMO CRUTEM4 data. See Figure 2. The adjustments to the GISS data create a slightly higher trend, roughly 0.01 deg C/decade. Not too surprisingly, the adjustments to the NOAA/NCEI land surface temperature data show the greatest change in reported warming rates from 1950 to 2015, over 0.02 deg C/decade.
NOTE: Keep in mind that NOAA failed to correct for the 1945 discontinuity and trailing biases in their new “pause-buster” ERSST.v4 sea surface temperature data. That failure on NOAA’s part drastically increases the warming rate of that dataset since 1950 compared to the sea surface temperature dataset that has been adjusted for the discontinuity and trailing biases, the HADSST3 data. See the discussion of Figure 2 in the post here and the post Busting (or not) the mid-20th century global-warming hiatus, which was also cross posted at Judith Curry’s ClimateEtc here and at WattsUpWithThat here. It appears as though NOAA is trying to minimize the mid-20th Century slowdown in global warming with land surface temperature data as well (every little bit helps), so they can claim global warming was continuous since 1950. [End note.]
TREND COMPARISONS FOR 1975 TO 2015
1975 is a commonly used breakpoint between the mid-20th Century slowdown and the recent warming period in global land+ocean data. So we’ll use that as the start of our next period for the “raw” versus “adjusted” land surface air temperature comparisons. See Figure 3. The warming rates of the “raw” GHCN data and all four of the “adjusted” datasets are basically the same for the period of 1975 to 2015. Curiously, all but the NOAA data show very slight decreases in reported trends with the adjustments over this period.
TREND COMPARISONS FOR 1998 TO 2015
1998 is commonly used as the start year for the slowdown in global warming. Figure 4 compares the “raw” and “adjusted” global land surface air temperature trends during this short-term period of 1998 to 2015. The UKMO CRUTEM4 data shows basically the same warming rate as the “raw” data. The adjustments to the GISS data created only a minor increase, about 0.01 deg C/decade. The Berkeley Earth and the NOAA data show the adjustments added about 0.02 deg C/decade to the reported warming rate.
The real curiosity is that the adjustments suppressed the El Niño-related upticks in 2015…well, not so much for the NOAA data as shown in the next graph.
LONG-TERM DIFFERENCES BETWEEN THE “RAW” AND “ADJUSTED” DATA
For those interested, Figure 5 presents the differences between the “raw” GHCN data and the “adjusted” data from Berkeley Earth, GISS, NOAA and UKMO. For this presentation, the data were referenced to the base years of 1880-1909 before the “raw” data were subtracted from the “adjusted” data. The early base years were used to provide a clearer illustration of the extent of the adjustments to the long-term data. The top graph includes the annual differences, and the bottom graph shows the annual differences smoothed with 5-year running-mean filters to reduce the annual volatility.
IMPACTS OF ADJUSTMENTS ON REPORTED LONG-TERM GLOBAL WARMING
It has recently become fashionable for alarmists to shift the data so that they can show how much global warming has occurred since “preindustrial” times. Unfortunately, most land surface air temperature datasets start well after “preindustrial” times, which, logically, are said to exist prior to the industrial revolution starting in the mid-1700s. So the best we can do is shift the data so that their linear trends align with zero at 1880. This also allows us to compare the “raw” and “adjusted” increases in global land surface air temperatures based on the linear trends. See Figure 6.
At the bottom of the illustration, I’ve listed the linear-trend-based changes in global land surface air temperatures from 1880 to 2015. The adjustments to the Berkeley Earth added about 0.35 deg C to the reported warming. For the GISS and NOAA/NCEI data, the adjustments increased the warming by roughly 0.3 deg C. The adjustments to the UKMO CRUTEM4 data only increased the reported warming since 1880 about 0.18 deg C.
As a reminder, we illustrated the decreases in long-term global warming that resulted from the adjustments to sea surface temperature data in the post here. In Figure 7, I’ve shifted the “raw” ICOADS sea surface temperature data and the sea surface temperature data used in the global land+ocean datasets so that their trend lines zero at 1880. Because the ocean surfaces on Earth cover more than twice the land surfaces (roughly 70% ocean versus 30% land), the upward trend adjustments to the land-based surface temperature data only offset a portion of the downward trend adjustments to the ocean surface temperature data.
Curiously, the trends of the “raw” sea surface temperature (Figure 7) and land surface air temperature (Figure 6) data are basically the same for the period of 1880 to 2015.
ONE MORE CURIOSITY
We mentioned above how the industrial revolution began in the mid-1700s. It just so happens that the Berkeley Earth land surface temperature data extend back to 1753. See Figure 8.
Important: the “global” land surface temperature data before 1880 have to be taken with a larger dose of salt than the data after 1880. To put that into perspective, the “global” land surface air temperature data in 1753 basically only include Europe, the eastern United States and eastern Canada. See the map here. Then again, Berkeley Earth presents it as “global”. With that in mind…
Like Figure 6, I’ve shifted the data so that the trend line equals zero at the first year of the data…in this case 1753. Based on the linear trends, the change in “global” land surface temperatures from 1753 to 2015 is less than the change in global land surface temperatures from 1880 to 2015.
Just a curiosity. The lack of data throughout the rest of the continental land masses in the early part of the data, and the resulting uncertainties, prevent us from knowing what land surface air temperatures were globally. Also, if you’re wondering, the ICOADS sea surface temperature data do not extend back that far, and, if memory serves, the very early ICOADS data are limited to parts of the North Atlantic.
The title question of the post was Do the Adjustments to Land Surface Temperature Data Increase the Reported Global Warming Rate?
As illustrated and discussed in this post, the answer is yes for the long-term land surface air temperature data.
For shorter-term periods (starting in 1950, 1975 and 1998), whether or not the adjustments have noticeable impacts land surface air temperature trends depends on the dataset and time period. Then again, as shown in the post here, the adjustments to the sea surface temperature data over those shorter timespans can increase the warming rates noticeably.
Again, many thanks to Zeke Hausfather for supplying the update for the unadjusted global GHCN land surface air temperature anomaly data.
For the next post, we’re going to compare the “raw” versus “adjusted” land+ocean surface temperature data for the NOAA, GISS and UKMO data. In other words, we’re going to merge the results from parts 1 and 2 of this series.