84
slope of 0.054 ± 0.05 °C/decade over this 15-year period, about a factor of two less
than the modeled slope of 0.108 ± 0.03 °C/decade. These two slopes do agree within
their respective uncertainties and, as is visually apparent, the ~155-year long simu-
lation does capture the essence of the observed variations reported by CRU over the
time period of the so-called hiatus. Nonetheless, the slopes disagree by a factor of
2, lending credence to the idea that some change in the climate system not picked
up by the EM-GC approach could be responsible for a gap between the modeled and
measured ΔT between 1998 and 2012.
a
b
c
d
Fig. 2.14 Observed and EM-GC simulated ΔT, 1995–2016. Top rung of a typical ladder plot,
comparing EM-GC modeled (red) and observed (grey) ΔT. Also shown are linear fits to the modeled
(red dashed) and measured (black) time series of ΔT, considering monthly values from the start of
1998 to the end of 2012. The slope and standard error of each slope are also recorded. (a) ΔT from
CRU was used (Jones et al. 2012 ); (b) ΔT from GISS (Hansen et al. 2010 ); (c) ΔT from NCEI (Karl
et al. 2015 ); (d) ΔT from the CRU Hybrid adjustment of Cowtan and Way ( 2014 ). The linear fits to
modeled ΔT for NCEI and CRU-H lie right on top of the respective fits to measured ΔT
2 Forecasting Global Warming