Satellite Era Global Temperature Anomaly Comparisons

It is interesting and instructive to compare different independent estimates of global temperature anomalies.  The largest amount of data available for input to global temperature estimates has occurred with the advent of satellite weather surveillance and global weather forecast models in the late 1970’s.  Prior to that time, data coverage was much more limited, especially over the oceans that make up about 71% of the earth’s surface.  Therefore the older estimates are likely to be less accurate and thus more uncertain.  Consequently, this comparison will focus on the period from 1979 to present where we have more data.

There are three main relatively independent sources of global surface temperature anomaly estimates available, derived from:  the Global Historical Climate Network (and sometimes including additional land stations) coupled with sea surface temperature measurements (GHCN), global weather forecast model input data, and satellite estimates of lower tropospheric temperatures.  There are multiple groups that compile and publish estimates from these sources but for simplicity, estimates from four of these groups are presented here.  For the GHCN related estimates, this analysis uses the US National Center for Environmental Information (NCEI) estimates and the Berkeley Earth Surface Temperature (BEST) estimates.  These are compared to the satellite estimates from the University of Alabama Huntsville (UAH) and global forecast system (GFS) estimates provided by the University of Maine (UM) Climate Change Institute (CCI).

Monthly global temperature anomaly estimates from all four groups representing three independent sources are presented in Figure 1 for the period 1979 through September 2015 and normalized to the 30-year climatic reference period 1981 through 2010.  The most widely publicized of these estimates are from NCEI and are highlighted in bold red.  In general, there is a fairly close agreement, typically within plus or minus 0.2 degrees Celsius (C) for most months.  The general trend over the entire 37 year period is upward by about 0.4C.  The variability between these estimates hints at the uncertainty but does not fully characterize it.  My best guess is that the uncertainty of estimates during this period is at least plus or minus 0.3C to 0.5C, as described in more detail here:
Uncertainty in Global Temperature Assessments

Considering the size of the uncertainty compared to the size of the trend, my opinion is that we can only say with low confidence that there may be a small rise in the global temperature during this period.

Global monthly temperature anomaly estimate comparisons for 1979-2015

Figure 1. Global monthly temperature anomaly estimate comparisons for 1979-2015

To allow more direct comparisons, Figures 2 through 5 present estimates for two groups at a time.  The GHCN related estimates from BEST and NCEI are compared in Figure 2.  As might be expected, these estimates show a striking similarity, although there is a hint of slightly larger monthly variations in the BEST estimates.  They both show a very similar trend.

Global Temperature Anomaly Estimates 1979-2015 BEST NCEI

Figure 2. Global temperature anomaly estimates 1979-2015 for GHCN BEST and GHCN NCEI

Figure 3 compares the GFS UM CCI estimates with the GHCN NCEI estimates.  For most of the period the two estimates are close, with the exception of two periods where the GFS UM CCI estimates are consistently slightly higher for 1980-1981 and 2002-2007 and for the most recent period since about 2010 where the GHCN NCEI estimates have suddenly departed substantially higher than the GFS UM CCI estimates by a fairly constant offset near 0.2C.  I am not aware of any major changes in the methodology or data inputs for the GFS UM CCI, but the GHCN NCEI estimates are constantly being revised and adjusted.  It appears that recent adjustments may have introduced a substantial high bias since 2010, especially considering that the GFS UM CCI and UAH satellite estimates show a fairly good match during this period.

Global Temperature Anomaly Estimates 1979-2015 for GFS UM CCI and GHCN NCEI

Figure 3. Global temperature anomaly estimates 1979-2015 for GFS UM CCI and GHCN NCEI

The UAH satellite estimates show the same general trend as the GHCN NCEI estimates, but a larger monthly variability as shown in Figure 4.  Considering that the UAH satellite estimates are for the lower troposphere whereas the GHCN NCEI estimates are for the surface, this comparison is easily within the previously mentioned uncertainty range.  This comparison also shows a hint of a possible high bias in the GHCN NCEI estimates since 2010.

Global Temperature Anomaly Estimates 1979-2015 for Satellite UAH and GHCN NCEI

Figure 4. Global temperature anomaly estimates 1979-2015 for Satellite UAH and GHCN NCEI

The UAH satellite estimates seem to subjectively show a slightly better comparison to the GFS UM CCI estimates as seen in Figure 5 than compared with the GHCN NCEI estimates in Figure 4.  The main deviations seem to be associated with El Niño and La Niña events.  The UAH satellite estimates are much higher during the very strong 1998 El Niño and slightly higher during weaker El Niños in 1983, 1987, 1991, and 2010 and somewhat lower during the 2007-2008 La Niña period.  The UAH satellite and GFS UM CCI estimates show a good agreement since 2010, unlike the suspect GHCN NCEI estimates.

Global Temperature Anomaly Estimates 1979-2015 for GFS UM CCI and Satellite UAH

Figure 5. Global temperature anomaly estimates 1979-2015 for GFS UM CCI and Satellite UAH

These comparisons give me less confidence in the GHCN NCEI estimates, especially since 2010.  My feeling is that the GFS UM CCI estimates, which are based on data with much better spatial coverage than the GHCN related estimates, probably provide the most accurate surface temperature anomaly estimates for this period.  In my next post, I will compare the trends in these estimates across the full 37 year period and then in a following post trends for the current century so far 2001-2015.

Monthly global temperature anomaly sources:

NCEI
http://www.ncdc.noaa.gov/cag/time-series/global/globe/land_ocean/p12/12/1880-2015.csv

BEST
http://berkeleyearth.lbl.gov/auto/Global/Land_and_Ocean_complete.txt

UAH
http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tlt/tltglhmam_6.0beta3.txt

UM CCI
http://cci-reanalyzer.org/Reanalysis_monthly/output/tseries_1_0_0_12_0.csv

Advertisements

2 responses to “Satellite Era Global Temperature Anomaly Comparisons

  1. Comparison of the satellite series with the surface temperature datasets is pointless in and of itself; it’s an apples to oranges comparison. The satellite TLT product is equivalent to approximately 4500m in altitude. The surface temperature data is at (obviously) the surface.

    It makes sense to compare the surface sets to each other and the satellite series to each other, but one should not arbitrarily compare surface to TLT unless one has a model to convert TLT to a surface temperature equivalent.

    With surface temperature datasets anyone interested can basically download the data and compute the results themselves. The data is publicly available and one can adjust or not adjust per one’s preference. The surface sets are very robust to the adjustment method, with the raw land/ocean data showing the largest historical trend while the individual series are very consistent with each other.

    The satellite series is more problematic. I don’t know of any layman that has actually taken the raw satellite data and built their own version of the TLT series. While RSS and UAH now match each other quite well, this was not always the case – even though RSS intentionally uses much of the UAH methodology. RSS has made some of their computer code public – UAH has not.

    The weighting functions for the atmosphere probably cause the biggest headache. For TLT products the sensors *see* both the cooling stratosphere and the warming surface. Removing these two influences is at the heart of the weighting functions. Roy Spencer of UAH wrote earlier this year regarding the new UAH V6.0: “”The new LT trend of +0.114 C/decade (1979-2014) is 0.026 C/decade lower than the previous trend of +0.140 C/decade, but about 0.010 C/decade of that difference is due to lesser sensitivity of the new LT weighting function to direct surface emission by the land surface, which surface thermometer data suggests is warming more rapidly than the deep troposphere.” From this one can see that slight changes in the weighting functions can have a significant effect.

    In sum, when one wants to discuss surface temperature, it’s probably best to stick to surface datasets. The satellite series are probably best compared to radiosonde data and used for a better understanding of atmospheric dynamics.

    • Kevin, I agree that surface and satellite global temperature anomaly estimates are like apples and oranges. They are related like different kinds of fruit, but should not be expected to match in every detail. From what I understand, the satellite TLT estimates are probably more like proxies than direct measurements, although I have not studied their derivation in detail. But that is also somewhat true of estimates of global temperature based on near surface air plus water temperatures. Global temperature is a concept and not something we can measure directly. The main interest in trying to estimate a global temperature anomaly is in trying to assess long-term trends. However, in my view, the uncertainty of these various estimates is much larger than is usually advertised and the uncertainty is especially high prior to the satellite era and much higher in the 19th century. Consequently, I have little confidence in trends less than about 1C per 100 years of data. By the end of this century that confidence may improve to more like plus or minus 0.5C for the 21st century, but I will not likely live to see that day.

      Another difficulty with surface temperature is: what does it represent over land with huge variations in terrain and ground cover? Surface measurements are typically made at 1.5 to 2 meters above ground level in open areas away from buildings and tall vegetation. But what is the surface temperature for a skyscraper urban core environment, or in extensive dense forests with canopies of 10 to 50+ meters above ground level, or in extremely mountainous terrain with elevation differences of a thousand meters or more in the space of kilometers? All we can say about our surface land temperature measurements is that they represent the typical exposure and location of our measurement sites.

      In reality, our greatest concern is the air temperature that will affect human activities and especially agriculture. Warmer temperatures favor expanding agriculture to larger areas and colder temperatures contraction. So, in the sense of human impact, our monitors probably do provide a reasonable representation but still lack good spatial coverage in all populated areas. However, as far as a true global temperature estimate, our GHCN monitors are woefully inadequate. This is the primary reason I favor the GFS reanalysis approach which incorporates a much larger number of surface temperature measurement locations. Can you imagine estimating surface temperatures around the globe using only measurements from GHCN sites for initialization of the global weather models four times each day?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s