Despite the apparent consilience of physical evidence surrounding climate change, it is a natural impulse and scientific obligation, to ask if there are measurements of some part of the climate system that depart from the expected or refute model predictions of climate change. If so, then it is important to understand if and how such measurements call us to revise projections of the future or reexamine the consensus of scientific view of climate change.
To date, the retrieval of temperatures by satellites are the most significant datasets to challenge mainstream interpretation of global warming. Some experts have argued that satellite data offer a fundamental critique of our present understanding of climate change, because they show much less warming than climate model predictions, and may indicate the need to lower temperature projections of climate change. There is not a agreement, however, over whether satellite temperatures actually do refute the predictions of climate models; as alternative explanations for the apparent disagreement have been closely examined, other possible explanations have emerged.
Since the late 1970s, satellites have measured radiation emitted by the atmosphere. In the 1990s, scientists developed methods for using these measurements to estimate the temperatures. The resulting datasets represent the temperature over tall bands of the atmosphere, several miles in height, with lots of uniform horizontal detail, but with errors from drifts in satellite orbit and the calibration between subsequent satellites making measurements. That is distinct from surface measurements, which offer a high level of vertical detail (temperature at 2 meters above the ground or at the surface of the ocean) but with limited spatial coverage, quality, and measurement type. Both kinds of data require data processing and careful scrutiny.
Earlier, flawed, versions of satellite temperature records showed slight cooling of the atmosphere,[2] a result wildly against model expectation and surface measurements of warming. Subsequent revisions have made these measurements less discordant with other evidence of climate change, but there is still a fairly good body of evidence climate models show larger temperature increases than the satellites. Whether that difference is against expectation or indicates a problem with climate models or theory depends on statistical tests and physical interpretation.
Figure Q2 shows different visual comparisons of atmospheric temperatures from satellites and computer models. The casual reader could justifiably see a modeling failure or relative coherence between models and satellite data from interpreting these graphics in isolation. The way the different lines are plotted against each other yields very different qualitative interpretations. If the intent is to understand if the satellite measurements are vastly different from climate models, then quantitative analysis is necessary.
Two comparisons of atmospheric temperatures measured from satellites with computer model simulations. The chart on the left shows the average prediction of 102 computer model simulations against averages of satellite and weather balloon datasets. The chart on the right shows the average computer model in black with a grey confidence interval against the average of satellite data in purple. Dataset versions and details vary between the two charts, but the difference between the two shows how graphical design can influence perception. The chart on the left comes from Congressional testimony and was not formally peer reviewed, while the chart on the right is excerpted from a peer-reviewed journal article.
There are three research groups that produce satellite temperature records for the whole globe. On average, they show that climate models have warmed faster than observations in the lower atmosphere. However, they do not agree if that difference is statistically significant, given the range in warmings projected by climate models. Without strong agreement on statistical significance between different scientific interpretations of satellite data, it is hard to draw a strong conclusion that these measurements are an indictment of our understanding of climate change.
Even without statistical significance, different interpretations might explain faster warming in climate model projections. The list is similar to the set of factors that could influence the global warming slowdown: models being, on average, too sensitive to enhanced CO2 warming. Other candidates include the climate models misrepresenting, by accident of experimental design, the true solar intensity and volcanic activity of the early 21st century or natural variability (the slowdown in the early 21st century) that reduced warming in the real world, common errors in the satellite data products, or some combination of these things.
The search for measurements that run contrary to the standing view on climate change is ongoing, as a matter of scientific discipline and as we enter a time when the expected signals of climate change should be detectable above natural variability. Satellite data, in particular, have historically confronted climate science with questions about the true temperature trends in the atmosphere and the basic physical understanding predicting that increase. However, recent developments in the scientific literature and repeated studies have lead us to a better understanding of the flaws in early processing of satellite data and an increasing reconciliation with climate models and theory. Details and refinements will continue to be worked out, but do not appear to fundamentally challenge the picture of developed by climate change science.