The determination of the earth system climate sensitivity and associated multi-decadal climate projections is a scientific problem with important social and making. In this thesis, we develop a new approach that directly exploits the stochasticity of the internal variability and the linearity of the response to radiative forcing to make projections based on historical data. At long enough timescales, nonlinear deterministic general circulation models (GCMs) are similarly stochastic and are believed to exhibit linearity of the mean response. Our approach produces the forced temperature by convolution of a Green's function, or Climate Response Function (CRF), with the radiative forcing. The full space-time CRF required is too general to allow...