Recent Research at GFDL on Surface Temperature Trends and Simulations of Tropical Cyclone Activity in the Indian Ocean Region - Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change - page 51

(b)
Fig. 1: Time series of Indian Ocean mean sea surface temperature (SST) anomalies
from observations (HadCRUT4; black curves) in degrees Celsius. CMIP5 20 th century
historical runs are shown (orange and blue curves) in degrees Celsius. The “All Forcing”
historical runs in (a) include 79 experiments across 23 different CMIP5 models. The
red curves show the multi-model ensemble means, which were computed by weighting
each model equally (as opposed to each individual model run equally). The “Natural
Forcing Only” historical runs in (b) are based on seven CMIP5 models (25 runs) that
had runs extending through 2010. Each individual series has been re-centred so that
the mean value for the years 1881-1920 is zero. Model data were masked with the
observed spatially and temporally evolving missing data mask.
equal weight in the average regardless of the number of ensemble members
that it contributes to the sample. The observations show a clear warming signal
that is fairly consistent with the All Forcing runs, but clearly inconsistent with
Natural Forcing Only runs.
Model-observational consistency can be assessed at least qualitatively in
Fig. 1. However, Fig. 2 illustrates how the consistency can be assessed more
quantitatively for a set of seven CMIP5 models. As described in the caption,
we use model control runs to build error bars for the model simulated trends
(the trend values plotted in the figure are for trends with varying start dates,
but all ending in 2010). Here we assume that the internal variability is not
affected by the forcing so that the control run variability is still applicable to
confidence intervals for the forced runs. The black-shaded region shows an