Interhemispheric gradient of atmospheric radiocarbon reveals natural variability of Southern Ocean winds

Tree ring Delta C-14 data (Reimer et al., 2004; McCormac et al., 2004) indicate that atmospheric Delta C-14 varied on multi-decadal to centennial timescales, in both hemispheres, over the period between AD 950 and 1830. The Northern and Southern Hemispheric Delta C-14 records display similar variability, but from the data alone is it not clear whether these variations are driven by the production of C-14 in the stratosphere (Stuiver and Quay, 1980) or by perturbations to exchanges between carbon reservoirs (Siegenthaler et al., 1980). As the sea-air flux of (CO2)-C-14 has a clear maximum in the open ocean regions of the Southern Ocean, relatively modest perturbations to the winds over this region drive significant perturbations to the interhemispheric gradient. In this study, model simulations are used to show that Southern Ocean winds are likely a main driver of the observed variability in the interhemispheric gradient over AD 950-1830, and further, that this variability may be larger than the Southern Ocean wind trends that have been reported for recent decades (notably 1980-2004). This interpretation also implies that there may have been a significant weakening of the winds over the Southern Ocean within a few decades of AD 1375, associated with the transition between the Medieval Climate Anomaly and the Little Ice Age. The driving forces that could have produced such a shift in the winds at the Medieval Climate Anomaly to Little Ice Age transition remain unknown. Our process-focused suite of perturbation experiments with models raises the possibility that the current generation of coupled climate and earth system models may underestimate the natural background multi-decadal- to centennial-timescale variations in the winds over the Southern Ocean.

Improved conjugate gradient implementation for least squares support vector machines

As a promising method for pattern recognition and function estimation, least squares support vector machines (LS-SVM) express the training in terms of solving a linear system instead of a quadratic programming problem as for conventional support vector machines (SVM). In this paper, by using the information provided by the equality constraint, we transform the minimization problem with a single equality constraint in LS-SVM into an unconstrained minimization problem, then propose reduced formulations for LS-SVM. By introducing this transformation, the times of using conjugate gradient (CG) method, which is a greatly time-consuming step in obtaining the numerical solution, are reduced to one instead of two as proposed by Suykens et al. (1999). The comparison on computational speed of our method with the CG method proposed by Suykens et al. and the first order and second order SMO methods on several benchmark data sets shows a reduction of training time by up to 44%. (C) 2011 Elsevier B.V. All rights reserved.