Long-memory effects in linear response models of Earth's temperature and implications for future global warming

Abstract

A linearized energy-balance model for global temperature is formulated, featuring a scale-invariant longrange memory (LRM) response and stochastic forcing representing the influence on the ocean heat reservoir from atmospheric weather systems. The model is parameterized by an effective response strength, the stochastic forcing strength, and the memory exponent. The instrumental global surface temperature record and the deterministic component of the forcing are used to estimate these parameters by means of the maximumlikelihood method. The residual obtained by subtracting the deterministic solution from the observed record is analyzed as a noise process and shown to be consistent with a long-memory time series model and inconsistent with a short-memory model. By decomposing the forcing record in contributions from solar, volcanic, and anthropogenic activity one can estimate the contribution of each to twentieth-century global warming. The LRM model is applied with a reconstruction of the forcing for the last millennium to predict the large-scale features of Northern Hemisphere temperature reconstructions, and the analysis of the residual also clearly favors the LRM model on millennium time scale. The decomposition of the forcing shows that volcanic aerosols give a considerably greater contribution to the cooling during the Little Ice Age than the reduction in solar irradiance associated with the Maunder Minimum in solar activity. The LRM model implies a transient climate response in agreement with IPCC projections, but the stronger response on longer time scales suggests replacing the notion of equilibrium climate sensitivity by a time scale–dependent sensitivity.