No Causation Without Correlation (FF↔GMT)


Previous posts addressed the claim that fossil fuels are driving global warming. This post updates that analysis with the latest (2018) numbers from BP Statistics and compares World Fossil Fuel Consumption (WFFC) with three estimates of Global Mean Temperature (GMT). More on both these variables below.


2018 statistics are now available from BP for international consumption of Primary Energy sources. 2019 Statistical Review of World Energy. 

The reporting categories are:
Natural Gas
Renewables (other than hydro)

This analysis combines the first three, Oil, Gas, and Coal for total fossil fuel consumption world wide. The chart below shows the patterns for WFFC compared to world consumption of Primary Energy from 1965 through 2018.

The graph shows that Primary Energy consumption has grown continuously for more than 5 decades. Over that period oil, gas and coal (sometimes termed “Thermal”) averaged 89% of PE consumed, ranging from 94% in 1965 to 85% in 2018.  MToe is millions of tons of oil equivalents.

Global Mean Temperatures

Everyone acknowledges that GMT is a fiction since temperature is an intrinsic property of objects, and varies dramatically over time and over the surface of the earth. No place on earth determines “average” temperature for the globe. Yet for the purpose of detecting change in temperature, major climate data sets estimate GMT and report anomalies from it.

UAH record consists of satellite era global temperature estimates for the lower troposphere, a layer of air from 0 to 4km above the surface. HadSST estimates sea surface temperatures from oceans covering 71% of the planet. HADCRUT combines HadSST estimates with records from land stations whose elevations range up to 6km above sea level.

Both GISS LOTI (land and ocean) and HADCRUT4 (land and ocean) use 14.0 Celsius as the climate normal, so I will add that number back into the anomalies. This is done not claiming any validity other than to achieve a reasonable measure of magnitude regarding the observed fluctuations.

No doubt global sea surface temperatures are typically higher than 14C, more like 17 or 18C, and of course warmer in the tropics and colder at higher latitudes. Likewise, the lapse rate in the atmosphere means that air temperatures both from satellites and elevated land stations will range colder than 14C. Still, that climate normal is a generally accepted indicator of GMT.

Correlations of GMT and WFFC

The next graph compares WFFC to GMT estimates over the 5+ decades from 1965 to 2018 from HADCRUT4, which includes HadSST3.

Over the last five decades the increase in fossil fuel consumption is dramatic and monotonic, steadily increasing by 234% from 3.5B to 11.7B oil equivalent tons.  Meanwhile the GMT record from Hadcrut shows multiple ups and downs with an accumulated rise of 0.74C over 53 years, 5% of the starting value.

The second graph compares WFFC to GMT estimates from UAH6, and HadSST3 for the satellite era from 1979 to 2018, a period of 39 years.

In the satellite era WFFC has increased at a compounded rate of nearly 2% per year, for a total increase of 91% since 1979. At the same time, SST warming amounted to 0.42C, or 3% of the starting value.  UAH warming was 0.44C, or 3% up from 1979.  The temperature compounded rate of change is 0.1% per year, an order of magnitude less.  Even more obvious is the 1998 El Nino peak and flat GMT since.


The climate alarmist/activist claim is straight forward: Burning fossil fuels makes measured temperatures warmer. The Paris Accord further asserts that by reducing human use of fossil fuels, further warming can be prevented.  Those claims do not bear up under scrutiny.

It is enough for simple minds to see that two time series are both rising and to think that one must be causing the other. But both scientific and legal methods assert causation only when the two variables are both strongly and consistently aligned. The above shows a weak and inconsistent linkage between WFFC and GMT.

Going further back in history shows even weaker correlation between fossil fuels consumption and global temperature estimates:


Figure 5.1. Comparative dynamics of the World Fuel Consumption (WFC) and Global Surface Air Temperature Anomaly (ΔT), 1861-2000. The thin dashed line represents annual ΔT, the bold line—its 13-year smoothing, and the line constructed from rectangles—WFC (in millions of tons of nominal fuel) (Klyashtorin and Lyubushin, 2003). Source: Frolov et al. 2009

In legal terms, as long as there is another equally or more likely explanation for the set of facts, the claimed causation is unproven. The more likely explanation is that global temperatures vary due to oceanic and solar cycles. The proof is clearly and thoroughly set forward in the post Quantifying Natural Climate Change.

Background context for today’s post is at Claim: Fossil Fuels Cause Global Warming.


  1. Bob Webster · July 15, 2019

    The lack of correlation between changing atmospheric CO2 and changing global average surface temperature (GAST) is even more instructive.

    Theory claims “changing atmospheric CO2 will produce corresponding changes in GAST” (growing CO2 warms, declining CO2 cools, unchanging CO2 produces temperature stasis). But the records contradict the narrative.

    Correlation between changing atmospheric CO2 and changing GAST is even lower than correlation of the history of the two.

    So the theorized relationship (described above) is refuted by the records. Any theory that defies observed evidence is junk (refuted) and must be rejected (in accordance with The Scientific Method).

    Politicians banking on “Climate Change” to give them a boost in 2020 are in for a big chill come election night.

    Liked by 1 person

    • Ron Clutz · July 15, 2019

      Bob, what you say is true. However, the relation between CO2 and GMT is confounded by temperature being a driver of CO2. At least WFFC and GMT are independent variables. As well, WFFC is claimed to be the source of all rising CO2, so the failure of that correlation invalidates policies aimed at reducing WFFC.


      • Bob Webster · July 15, 2019

        Ron, Your analysis is good and I certainly do not take any issue with it. However, because the IPCC vastly overstates the impact WFFC has on global average CO2, I set out to deliberately avoid that issue by examining the relationship between total CO2 (including whatever fossil fuel use contributes) and GAST (or GMT).

        The record shows the relationship to be one of pure chance (50% of the time, annual global average CO2 change and annual GAST change are consistent with each other; 50% of the time they are inconsistent with each other). The flip of a fair coin does a better job projecting future climate than does the costly array of CMIP5 climate change models based on an invalid theory. Don’t blame the modelers… blame the theory.

        Since 1880, the observed records show correlation of changing CO2 and changing GAST is generally in the area of 0.1 to 0.2 and always less than 0.3. Consequently, any theory that claims GAST changes are caused by changing global average CO2 is clearly invalid, regardless of what is causing CO2 to change. No correlation, no causation.

        The Greenland GISP2 record for the Holocene interglacial shows the opposite of what climate change theory predicts… peak temperatures of the past 10,700 years occur when the lowest atmospheric CO2 is observed in the record and the lowest temperatures are found when the highest atmospheric CO2 is observed in the record.

        No matter how we look at it, there is no correlation between atmospheric CO2 and climate change and clearly there is no correlation between WFFC and climate change.

        Certainly, if atmospheric CO2 is uncorrelated with GAST, then the amount of atmospheric CO2 contributed by fossil fuel use must be entirely irrelevant, as you’ve observed.


    • paulaubrin · July 16, 2019

      “The lack of correlation between changing atmospheric CO2 and changing global average surface temperature (GAST) is even more instructive.”
      Indeed, when you compare variations, you get rid of the zero frequency component, which is often spurious. Then it is possible to compute the correlation coefficient of variations for various lag times.


  2. Hifast · July 15, 2019

    Reblogged this on Climate Collections.


Leave a Reply to Hifast Cancel reply

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

You are commenting using your 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 )

Connecting to %s