Part II. How do we know that the earth is warming?

In my previous post on global warming, I described the greenhouse effect—the mechanism behind global warming. The point I made is that the greenhouse effect is based on fundamental laws of physics. It must occur! As levels of greenhouse gases increase, global temperatures must also increase.

In this post we look at the data that show temperatures are rising in parallel with the rising level of carbon dioxide (CO2) in the atmosphere. As in my previous post, I provide citations so that you can trace the data back to peer-reviewed scientific publications. (For the importance of citing sources, see my post on How Do You Know That?.) The endnotes also provide references to analysis of the data that is more thorough than can be presented here.

The Intergovernmental Panel on Climate Change (IPCC) concluded in its most recent report: “Human activities, principally through emissions of greenhouse gases, have unequivocally caused global warming[1]. This is a strong statement from the usually reserved and cautious IPCC.

Let’s break down how they reached this conclusion. I will try to answer questions such as: (1) How do we know temperatures are rising and how reliable is that data? (2) What do we know about global temperatures in the past? (3) Can we confidently conclude that rising temperatures result from anthropogenic (human-caused) greenhouse gas emissions?

1. How do we know that temperatures are, in fact, rising?

Recorded temperatures

Reliable recorded temperatures go back to about 1850, so the first thing we can do is look at those records. Weather stations record air temperatures close to the land surface. Ships and buoys make temperature measurements in the oceans. Together these readings give us some of the best temperature data available. The global mean surface temperature (GMST) is an average of land-surface air temperatures and sea surface temperatures around the globe, including both northern and southern hemispheres.

The figure below (Figure 1) shows three different compilations [2] of the global average surface temperature. These data sets include millions of measurements [3]. The three compilations agree closely and each shows that the temperature has been rising, especially since around 1970. The decades-long rise in temperatures is clear even though from year to year the temperature may fluctuate up or down. (The fluctuations arise because factors such as El Niño and La Niña affect the global temperature from year to year.)

Figure 1. Global Mean Surface Temperature (GMST) based on temperature measurements since 1850, plotted as the “temperature anomaly,” the yearly global average temperature relative to the average temperature between 1850 and 1900 [4]. The GMST is the global average including both land surface air temperatures (as measured by weather stations) and sea surface temperatures (as measured by buoys and ships). Note the increase since 1970 of about 0.2°C per decade. The gray shading shows the 2.5% to 97.5% confidence range.

How reliable are the direct temperature records?

Scientists put a lot of effort into assessing uncertainties in their data, and the records plotted in Figure 1 are no exception. Confidence intervals from statistical analysis of the data tell us about uncertainty levels. In Figure 1 the gray shading shows the interval between the 2.5% and 97.5% confidence limits for the HadCRUT data set. That means the probability is 95% that the true value lies within the shaded region for each data point, while the chance is only 2.5% chance (1 in 40) that the true value is lower and 2.5% chance that the true value is higher. Note that over the past 50 years, the gray region is so small that it is hardly visible on the plot. So the data are very reliable. The chances are virtually nil that the observed temperature rise over the past 50 years is not real.

Bringing together temperatures recorded from thousands of weather stations, buoys, and ships leads to high confidence in the data but also raises questions about how to merge the multiple temperature readings. Some thermometers are more accurate than others, and some have systematic biases in one direction or another, reading consistently a little high or a little low. Or new thermometer may replace an old one. Such factors need to be (and are) taken into account. Temperature differences are still accurate, even if the thermometer consistently reads a little high or a little low. If a given thermometer consistently reads 0.1°C high, it can still accurately measure a temperature change of a fraction of a degree. So averaging temperature differences from various stations yields an accurate result.

Other issues also arise. For example, if an urban area spreads around a weather station over the decades, maybe the temperature at the station increases because of its surroundings, not because of global warming but because of the “urban heat-island effect.” This possibility (and others) have been thoroughly studied and taken into account in the data analysis.

Confirmation by independent analysis

In 2010, several scientists, including Richard Muller, a physics professor at the University of California Berkeley, started a nonprofit organization called Berkeley Earth. In response to criticisms by skeptics, Muller and colleagues set out systematically to reanalyze data to exclude factors such as the urban heat island effect mentioned above. Their analysis agrees closely with other analyses, as we can see in Figure 1 above, which includes Berkeley Earth’s result. Funded by organizations including the conservative Charles G. Koch Foundation, Berkeley Earth set out to find the “truth” about global warming, only to find that climate scientists were already telling the truth.

So yes, temperatures are rising. For a particularly vivid illustration of rising temperatures over the past century see this graph (Figure 2) from Berkeley Earth.

Figure 2. Monthly globally averaged temperatures by month from 1921 to 2023 plotted as the difference from the 1850-1900 average. The general trend is towards warmer monthly temperatures with time. Earlier years in blue are at lower temperatures and later years, yellow and red, at higher temperatures. The year 2023 was globally the hottest year on record for over 100 years (and quite possibly much longer. Graphic from Berkeley Earth licensed under Creative Commons BY-NC 4.0 International for non-commercial use only.

2. Can we compare current temperatures to historical temperatures over a longer time span?

Temperature proxies

Thermometers were invented in the 1600s, but reliable thermometer readings go back only to around 1850. That’s not that long ago! But there are indirect ways to deduce what temperatures were before then. These methods rely on temperature “proxies.” A proxy is a natural record that is correlated with temperature. Examples are tree rings (by analysis of width, density, isotopic composition), ice cores (which trap dust and air from the past), sediments (in lakes and oceans), speleothems (formations such as stalactites and stalagmites in caves), and corals (the density and oxygen isotope composition of coral skeletons). For example, the ratio of two isotopes of oxygen, oxygen-18 (18O) and oxygen-16 (16O, the most abundant isotope of oxygen) can be read as a natural thermometer [6]. Climate scientists analyze oxygen from ice cores, corals, and calcite shells of marine organisms such as foraminifera in ocean sediment in this way.

Figure 3 below shows the reconstructed temperature record of global surface temperatures over the past 2000 years, with direct measurements (in green) superimposed on the reconstruction since 1850. The figure includes for comparison the atmospheric CO2 concentrations over the same time period (described in my previous post).

Figure 3. Atmospheric CO2 concentrations (top) and global mean surface temperatures (bottom) since 1 AD [7]. GMST is reconstructed from a compilation of temperature proxies [5]. The blue shading shows the 5% to 95% confidence region. Direct observations from 1850 on (green) are superimposed on the reconstructed temperatures.

How reliable are temperature proxies?

Paleoclimatologists compile data from hundreds of proxies to estimate past temperatures [5]. Proxies have been tested against each other and against direct temperature measurements from the time period since 1850. The agreement is strong. In Figure 3, the blue proxy curve follows the average of the direct measurements shown in green, reproducing the temperature rise since 1970.

Still, uncertainties become larger as we look back in time via temperature proxies. The blue shaded region in Figure 3 shows the region between the 5% and 95% confidence limits. That means there is a 90% chance that the true temperature for each data point is within the blue shading. So it is nearly certain that temperatures over the past 50 years are higher than they have been for over 2000 years. Taking proxy data back even farther, the latest IPCC report concludes that average temperatures in the past decade are the highest going back around at least 6,500 years [8]—since the beginnings of agriculture.

3. How do we know that the rise in temperature is a result of increasing greenhouse gas emissions rather than natural cycles?

The causal link between CO2 and temperature

Figure 3 shows that the temperature rise tracks the rise in CO2 level. But, just because two things happen at the same time does not prove that one causes the other.

There are several indicators of cause. First, we know the mechanism that connects greenhouse gases to rising temperatures. (This was the topic of my previous post.) The greenhouse effect is basic science: increasing greenhouse gases must cause increasing temperature.

Second, the temperature rise has been rapid and unprecedented (again see the temperature record in Figure 3). In fact, global temperature have increased more rapidly over the last 50 years than any other 50-year period in at least 2000 years [9].

Climate models

Third, the rise in temperature agrees closely with climate models. (A climate model is a computer simulation that takes factors such as solar radiation, cloud cover, atmospheric temperatures, ocean temperatures, land cover, ice cover, greenhouse gas concentrations, etc., as input for computations based on the physics and chemistry of the atmosphere, land, and oceans.) Climate scientists use some of the world’s most powerful computers to predict future climate conditions.

We cannot do a control experiment where we rerun the last 50 years on earth without the increase in greenhouse gases, but we can make that comparison with computer models. A comparison of computer models run with or without the increase in greenhouse gases shows that the temperature rise that has occurred results only when greenhouse gases are included [10]. That means the temperature rise is not the result of, say, solar cycles, changes in earth’s orbit, or natural variability. (All of these possibilities have been extensively investigated.)

All of this evidence leads to the conclusion that the globe is warming and that the cause is unequivocally [1] greenhouse gas emissions by humans.

Now there is one more important question to address: So what if the temperature rises by a few degrees. What’s the big deal? A rise in temperature of 1° to 2°C (2 to 4°F) may not sound like much. After all, temperatures vary by a lot more than that during the day every day. But when we are talking about global averages, it is a big deal! It’s more like having a high fever. To see why, we need to look at climate predictions, and that will be the topic of my next post.


[1] The Intergovernmental Panel on Climate Change (IPCC) is an international panel sponsored by the United Nations. Its task is to assess and summarize the scientific data on climate change. Their latest report is the Sixth Assessment Report (AR6). The statement quoted comes from AR6 Synthesis Report: Climate Change 2023, pg. 4, available online at

[2] One compilation is by the National Oceanic and Atmospheric Administration (NOAA), one by the Meteorological Office (Met Office) Hadley Centre in the United Kingdom, and one from Berkeley Earth, an independent, non-profit organization.

[3] See IPCC, 2021: Climate Change: The Physical Science Basis (IPCC WG1 2021), Section, Table 2.3, and Figure 2.11.

[4] Different data sets report the temperature anomaly relative to averages from different time periods. For example, the Met HadCRUT5 data set is reported relative to the period 1960-1990, NOAA relative to 1901-2000, and Berkeley Earth relative to 1951-1980. Whatever the reference temperature is makes no difference to the changes observed relative to that temperature. Following the IPCC WG1 AR6 report, I have plotted the data relative to the period 1850-1900, and my plot closely resembles the plot in Figure 2.11c of the IPCC WG1 2021 report.

[5] The PAGES2k consortium reconstructs surface temperature over the past 2000 years based on multiple proxies. The results are peer-reviewed. See Neukom et al., Nature Geoscience 12, 643–649 (2019). For a reconstruction based on over 1000 proxy records going back almost 10,000 years, see Kaufman et al., Scientific Data 7, 201 (2020) and Figure 2.11a in IPCC WG1 2021 Chapter 2.

[6] Water containing 16O evaporates slightly more readily than water containing 18O, and since the evaporation rate depends on temperature, the ratio of the amounts of the two isotopes gives information about the temperature at the time when precipitation (rain or snow) occurred. Similarly, the ratio of the two isotopes in the calcite shells of sea creatures depends on ocean temperature. See John Houghton, Global Warming. The Complete Briefing, 5th edition (Cambridge University Press) pg. 81.

[7] The data for Figure 3, showing globally averaged surface temperatures, come from the Sixth Assessment Report of the Intergovernmental Panel on Climate change, Working Group 1, Chapter 2, Figure 2.4 and 2.11. As I noted in my previous post, the data come from the peer-reviewed scientific literature, and the report was itself peer-reviewed. This establishes the “scientific credentials” of the data. Data sources are available from the Supplementary Materials for that chapter. See  

[8] IPCC WG1 2021 report, pg. 290.

[9] IPCC, 2021: Summary for Policymakers. In: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V. et al. (eds.)], pg. 8.

[10] IPCC, 2021: Summary for Policymakers (see previous endnote), Fig. SPM.1.


Leave a Reply

Avatar placeholder

Your email address will not be published. Required fields are marked *