Statistical perspectives on stratospheric transport

Authors
Citation
Lc. Sparling, Statistical perspectives on stratospheric transport, REV GEOPHYS, 38(3), 2000, pp. 417-436
Citations number
47
Categorie Soggetti
Earth Sciences
Journal title
REVIEWS OF GEOPHYSICS
ISSN journal
87551209 → ACNP
Volume
38
Issue
3
Year of publication
2000
Pages
417 - 436
Database
ISI
SICI code
8755-1209(200008)38:3<417:SPOST>2.0.ZU;2-3
Abstract
Many long-lived stratospheric chemical constituents enter the stratosphere through the tropical tropopause, are transported throughout the stratospher e by the Brewer-Dobson circulation, and are photochemically destroyed in th e upper stratosphere. These chemical constituents, or "tracers," can be use d to track mixing and transport by the stratospheric winds. Much of our und erstanding about the stratospheric circulation is based on large-scale grad ients and other spatial features in tracer fields constructed from satellit e measurements. The point of view presented in this paper is different, but complementary, in that transport is described ill terms of tracer probabil ity distribution functions. The probability distribution function is comput ed from the measurements and is proportional to the area occupied by tracer values in a given range. The flavor of this paper is tutorial, and the ide as are illustrated with several examples of transport-related phenomena, an notated with remarks that summarize the main point or suggest new direction s. The examples illustrate how physically based statistical analysis can sh ed some light on aspects of stratospheric transport and dynamics that may n ot be obvious or quantifiable with other types of analyses. The dependence of the statistics on location and time is also shown to be important for pr actical problems related to statistical robustness and satellite sampling. An important motivation for the work presented here is the need for synthes is of the large and growing database of observations of the atmosphere and output generated by atmospheric models.