NASA scientist Roy Spencer recently posted on his Web site some startling graphs produced by John Christy, his colleague at the University of Alabama in Huntsville. The graph immediately below compares the linear-trend temperature projections of 73 climate models with the linear trend of observed temperatures for the bulk tropical atmosphere during 1979-2012.
The 73 models are part of the fifth phase of the Coupled Model Intercomparison Project (CMIP-5), a collaborative effort of 20+ modeling groups to inform the IPCC’s forthcoming Fifth Assessment Report (AR5). The Project’s three main objectives are to “evaluate how realistic the models are in simulating the recent past,” “provide projections of future climate change” out to 2035 and 2100, and “understand some of the factors responsible for differences in model outputs” such as different estimates of feedback effects.
Christy’s graph reveals what Spencer calls an “epic failure” of the models to match the actual behavior of the tropical atmosphere. Models that overestimate recent warming are likely to overestimate future warming as well.
Of course, observational systems may have biases and errors, but that is an implausible explanation for the mismatch. The observations come from two satellite and four radiosonde (weather balloon) datasets, which all independently give “virtually identical trends.”
What about the subset of U.S.-designed models — do they get the trend right? Nope. Take a gander at the next graph.
I continue to suspect that the main source of disagreement is that the models’ positive feedbacks are too strong . . . and possibly of even the wrong sign.
The lack of a tropical upper tropospheric hotspot in the observations is the main reason for the disconnect in the above plots, and as I have been pointing out this is probably rooted in differences in water vapor feedback. The models exhibit strongly positive water vapor feedback, which ends up causing a strong upper tropospheric warming response (the “hot spot”), while the observation’s lack of a hot spot would be consistent with little water vapor feedback.
Some visitors to Spencer’s site complained that “linear trends are not a good way to compare models to observations.” They miss the point. Linear plots show the most fundamental quantity the models are supposed to reveal — the long-term rate of heat accumulation in the tropical troposphere due to greenhouse gases. The graphs above are entirely legitimate.
Nonetheless, to oblige the critics, Spencer and Christy plotted five-year running averages for observations and projections during 1979-2012. The disconnect persists.
Seeing is believing, but things are not always what they seem. Skeptical Science, a Web site devoted to debunking global warming skepticism, asserts that Spencer’s claim about recent warming being only 50% of what the model consensus projects is “flat-out ridiculously wrong” (original emphasis). Observed warming has been “spot on consistent with climate model projections,” Skeptical Science contends. The evidence, supposedly, is in the graph below (click on it to activate the presentation).
Figure explanation: This animation compares the observed global temperature change since 1990 (black curve) to projections of global temperature change from the first four Intergovernmental Panel on Climate Change (IPCC) reports (red, pink, orange, green) and from various “climate contrarians” (blue, purple, green, gray dashed). The observations are given by the average of 3 primary global temperature datasets (NASA GISS, NOAA NCDC, and HadCRUT4). All of the IPCC projections have proven to be quite accurate, suggesting high reliability. The contrarian projections all underestimate the global warming substantially, and in fact they erroneously predict global cooling and are quite unreliable.
So who’s right: Spencer and Christy or Skeptical Science (SS)? The SS graph and commentary are misleading in two ways.
The period covered in the SS graph is a decade shorter than that covered by the Spencer-Christy graph and looks suspiciously like cherry-picking. By starting their graph in 1990, SS can use the Mt. Pinatubo-induced cold period of 1992-93 to tilt the trend to be more positive. The Spencer-Christy graph begins at the start of the satellite record — 1979 — providing a longer and more representative period.
More importantly, SS uses global surface temperature datasets, which do not accurately represent heat content in the bulk atmosphere. In contrast, Spencer and Christy use temperature data from the tropical troposphere — the place where the models project the strongest, least ambiguous, greenhouse warming signal.
As Christy explained in testimony last August, the popular surface datasets often touted as evidence of model validity are not reliable indicators of the greenhouse effect. Land use changes (urbanization, farming, deforestation) “disrupt the normal formation of the shallow, surface layer of cooler air during the night when TMin [daily low temperature] is measured.” Over time, TMin gets warmer, producing a trend easily mistaken for a global atmospheric phenomenon.
Surface temperatures are not a direct or reliable measure of bulk atmospheric heat content, which is one reason Christy has devoted much of his career to developing a satellite record of global temperatures. Satellite datasets “are not affected by these surface problems and more directly represent the heat content of the atmosphere.”
If the models accurately represented the atmosphere’s heat content, Christy told me by email, the warming rate of the tropical mid-troposphere (TMT) would be 1.4 times that of the surface. In reality, the TMT warming rate is only 0.6 times that of the surface. The models simply project too much warming in the bulk atmosphere.
Spencer sums up the situation thusly: “I frankly don’t see how the IPCC can keep claiming that the models are ‘not inconsistent with’ the observations. Any sane person can see otherwise.”