Journal of the Atmospheric Sciences, 54, 2241-2260, 1997
Adam H. Sobel1 and R. Alan Plumb
Center for Meteorology and Physical Oceanography, Massachusetts Institute of Technology, Cambridge, Massachusetts
Darryn W. Waugh2
Cooperative Research Center for Southern Hemisphere Meteorology, Monash University, Victoria, Australia
Existing quantitative calculations of material transport across the stratospheric polar vortex edge are difficult to interpret. This is because what is actually calculated has not been clearly shown to be irreversible transport, because of ambiguities inherent in defining the vortex edge, and (relatedly) because the uncertainties in the various sorts of calculations have not been quantified. We discuss some of the conceptual and technical difficulties involved in such calculations. These typically use a tracer coordinate, so that an air parcel's ``position'' is defined as a function of some tracer which it carries. We examine the sensitivity to noise of a method which has been used in several prior studies, which we call the ``contour crossing'' method. When contour crossing is implemented with no explicit threshold to discriminate noise from signal, a realistic amount of noise in the tracer data can cause apparent transports across the vortex edge in the range of ten percent to several tens of percent of the vortex area per month, even if the true transport is zero. Moreover, contour crossing does not discriminate between dynamically driven transport and that due to large-scale nonconservative effects acting upon the tracer used to define the coordinate. We introduce a new method, which we call the ``local gradient reversal'' method, for estimating the dynamically driven component of the transport. This method is conceptually somewhat similar to contour surgery, but applies to gridded fields rather than material contours. Like contour crossing, it can thus be used in conjunction with the reverse domain filling advection technique, while contour surgery is used with contour advection or contour dynamics. Local gradient reversal is shown to be less sensitive to noise than contour crossing.