Multi-Frequency Analysis of Simulated versus Observed Variability in Tropospheric Temperature

Abstract

Studies seeking to identify a human-caused global warming signal generally rely on climate model estimates of the “noise” of intrinsic natural variability. Assessing the reliability of these noise estimates is of critical importance. We evaluate here the statistical significance of differences between climate model and observational natural variability spectra for global-mean mid- to upper tropospheric temperature (TMT). We use TMT information from satellites and large multi-model ensembles of forced and unforced simulations. Our main goal is to explore the sensitivity of model-versus-data spectral comparisons to a wide range of subjective decisions. These include the choice of satellite and climate model TMT data sets, the method for separating signal and noise, the frequency range considered, and the statistical model used to represent observed natural variability. Of particular interest is the amplitude of the interdecadal noise against which an anthropogenic tropospheric warming signal must be detected. We find that on timescales of 5-20 years, observed TMT variability is (on average) overestimated by the last two generations of climate models participating in the Coupled Model Intercomparison Project. This result is relatively insensitive to different plausible analyst choices, enhancing confidence in previous claims of detectable anthropogenic warming of the troposphere and indicating that these claims may be conservative. A further key finding is that two commonly used statistical models of short-term and long-term memory have deficiencies in their ability to capture the complex shape of observed TMT spectra.

Robust detection of forced warming in the presence of potentially large climate variability

“…However, a key limitation of traditional D&A is that the robustness and estimated confidence levels depend on the ability of climate models to adequately simulate internal climate variability, particularly on longer multidecadal time scales ( 1 ). Comparisons between models and observations indicate that climate models show a plausible representation of global-scale temperature variability on interannual to centennial time scales ( 1 , 10 – 17 ), including the pattern representation of key modes of natural (internal) climate variability ( 18 , 19 ). Some studies infer a small role of multidecadal internal variability in the observed global temperature record ( 20 – 23 ) that is consistent with model simulated variability.…”

“…However, if the climate models used to obtain variability estimates from unforced control simulations were to systematically underestimate decadal-scale variability compared to the real world, then these D&A approaches would overestimate the signal-to-noise ratio, that is, the magnitude of the forced response relative to internal variability ( 1 , 3 , 17 , 54 ). The result would be a bias toward earlier detection times.…”

Robust detection of forced warming in the presence of potentially large climate variability

“…However, a key limitation of traditional D&A is that the robustness and estimated confidence levels depend on the ability of climate models to adequately simulate internal climate variability, particularly on longer multidecadal time scales (1). Comparisons between models and observations indicate that climate models show a plausible representation of global-scale temperature variability on interannual to centennial time scales (1,(10)(11)(12)(13)(14)(15)(16)(17), including the pattern representation of key modes of natural (internal) climate variability (18,19). Some studies infer a small role of multidecadal internal variability in the observed global temperature record (20)(21)(22)(23) that is consistent with model simulated variability.…”

“…However, if the climate models used to obtain variability estimates from unforced control simulations were to systematically underestimate decadal-scale variability compared to the real world, then these D&A approaches would overestimate the signal-to-noise ratio, that is, the magnitude of the forced response relative to internal variability (1,3,17,54). The result would be a bias toward earlier detection times.…”

“…The large spread in simulated internal variability across climate models implies that the observed 40-year GMT trend of 0.76°C for 1980-2019 would exceed the SD of internally generated variability of a set of "low-variability" models by about a factor of 5 or more (Fig. 1B), 17 corresponding to vanishingly small probabilities for the warming to be internally generated. However, the observed trend would exceed the SD of some "high-variability" models only by about a factor of 2, which would make internal variability a highly unlikely but not completely implausible explanation for a substantial fraction of the signal.…”