Figure 2

(A) distribution of observed decadal temperature trends (GISS) within the “hiatus” windows defined by the corpus of articles considered for this analysis (blue), compared to the distribution of all possible temperature trends from 1950 till 2012, the reference period used by the IPCC to establish the long-term warming trend (pink). (B) same distribution of temperature trends within the “hiatus” windows (blue) compared to the distribution of all possible temperature trends from 1964 till 2012 (pink). The year 1964 is the lower bound for the 95% confidence interval of a recent change-point analysis that sought to identify the onset of modern global warming. (C) same distribution of temperature trends within the “hiatus” windows (blue) compared to the distribution of all possible temperature trends from 1976 till 2012 (pink). The year 1976 is the upper bound for the 95% confidence interval of a recent change-point analysis that sought to identify the onset of modern global warming. In all panels, the distribution of all possible trends is obtained by computing all trends of a given duration from all possible years within the time period considered. The duration of trends is weighted by the propensity of presumed “hiatus” durations in the corpus. Thus, each 10-year trend is replicated 8 times (as 8 articles in the corpus presumed the “hiatus” to extend over 10 years), each 11-year trend 5 times and so on. See Table 1 for details of the distribution of presumed “hiatus” durations in the corpus. The vertical red lines in each panel represents the long-term trend (1951–2012) that was used by the IPCC in their Fifth Assessment Report as a benchmark for comparison with the “hiatus.” The solid line is for the GISS dataset⁴⁰ analyzed here and the dashed line is the same long-term trend using the UK Met Office’s HadCRUT4 data set³⁹ used by the IPCC.