...This is the only response I will make on Dr. Curry’s website.
1. We routinely consider “ANTHRO only” fingerprints – see, e.g., the discussion on page 7 of the Supplementary Material of the 2018 Santer et al. Science paper. That discussion explains why the “ANTHRO only” and HIST+8.5 fingerprints yield very similar results. In my opinion, it is not unreasonable to expect other scientists to read such background information, particularly since it is cited in the Nature Climate Change paper you are critiquing.
2. You suggest – incorrectly – that we never evaluate the adequacy of model-based estimates of internal variability. We routinely make such evaluations. Examples are given in Fig. S7 of the 2018 Santer et al. Science paper and in Figs. 9 and 10 of the 2011 Santer et al. JGR paper.
3. Readers of your blog post might infer that we are unconcerned with differences between modeled and observed tropospheric warming rates. That is not the case. Many of our publications have attempted to understand the causes of differences between simulated and observed warming rates in the early 21st century. In the 2017 Santer et al. Nature Geoscience paper, we find that a large error in model climate sensitivity – Dr. Christy’s preferred hypothesis for model-versus-data warming rate differences – does not explain the temporal structure of these differences.
4. The pattern comparison statistic we use in our “fingerprint” work is an uncentered spatial covariance. It is not a correlation.
5. Even if one ignores all pattern information and considers global-mean changes alone, the amplitude of observed tropospheric temperature changes remains large relative to model-based estimates of internal variability (see, e.g., Fig. 1E in the 2017 Santer et al. Scientific Reports paper). This holds even for University of Alabama tropospheric temperature data.
6. Whether we do or do not remove residual long-term drift from control run data has minimal impact on our results. We only detrend once (over the final 200 years of each control run). We do not detrend each L-year chunk we are processing when we estimate time-dependent S/N ratios.
7. It is true that “rebound” of tropospheric temperature from the cooling caused by Pinatubo contributes to observed warming over the satellite era. You neglect to mention that our group has studied volcanically induced “rebound” of tropospheric temperature since 2001 (see, e.g., Santer et al. 2001, JGR; Santer et al. 2014, Nature Geoscience). The rebound effect is relatively small over the entire 40-year satellite tropospheric temperature record. Additionally, it is impermissible to focus solely on “rebound” from the eruptions of El Chichon in 1982 and Pinatubo in 1991, and to ignore the cooling effects of early 21st century volcanic eruptions. The climate effects of post-2000 volcanic forcing have been studied in a number of publications (e.g., Solomon et al., Science, 2011; Ridley et al., GRL, 2014; Santer et al., GRL, 2015). The effect of these post-2000 eruptions is to reduce S/N ratios for analysis periods sampling temperature changes in the early 21st century.
8. In other fingerprint detection work, we have tested not only against model-based estimates of internal variability, but also against “total” natural variability (internally generated plus variability forced by changes in solar irradiance and volcanoes). See, e.g., the 2013 Santer et al. “vertical fingerprint” paper in PNAS. For changes in the vertical structure of atmospheric temperature, we can detect an anthropogenic fingerprint even against this larger “total” natural variability.
9. The control run distributions of noise trends are Gaussian (at least for tropospheric temperature).
Sincerely,
Ben Santer