Optimal detection of global warming using temperature profiles: A methodology

Optimal fingerprinting is applied to estimate the amount of time it would t ake to detect warming by increased concentrations of carbon dioxide in mont hly averages of temperature profiles over the Indian Ocean. A simple radiat ive-convective model is used to define the pattern of the warming signal, a nd the first 100 yr of the 1000yr control run of the Geophysical Fluid Dyna mics Laboratory atmospheric-oceanic global climate model is used to estimat e the natural variability of the upper-air temperatures. The signal is assu med to be the difference in two epochs of data, each epoch consisting of 12 consecutive months of monthly average temperature profiles. When the varia bilities of monthly averages are assumed independent of each other, the dif ference in August upper-air temperatures yields the strongest fingerprint, giving a time span for a one-sigma detection of 22 yr. When correlations of natural variability between months are considered, the one-sigma detection time is 10 yr. If only an annual average profile is used, the timescale fo r one-sigma detection increases to 14 yr. These timescales depend on subjec tive judgments of the details of the model-predicted pattern of global warm ing. In general, using upper-air temperatures adds approximately two indepe ndent pieces of information in detecting global warming for every surface-a ir temperature measurement, most likely due to the expected overall pattern of tropospheric warming-stratospheric cooling. Finally, testing climate mo dels with data must be undertaken in order to understand the uncertainties in model-predicted global warming patterns and the predictive capability of models in general.