Table of Contents * Previous Chapter * Next Chapter
The implicit time-stepping scheme described in Stott and Harwood (Annales Geophysicae, 11, 377-388, 1993) has been improved by reassessing the treatment of families and improving the convergence properties of the scheme. It has been christened IMPACT (IMPlicit Algorithm for Chemical Time-stepping). The two figures show how IMPACT performs when used for the stratospheric chemistry scheme designed by Glenn Carver and Sarah Goodchild using the ASAD package. This chemical scheme includes a detailed representation of chlorine, bromine and hydrogen species.
The solutions calculated by the IMPACT scheme are the SOLID lines for each constituent. IMPACT saves computing time by avoiding the inversion of matrices so it is interesting to compare IMPACT with a version of the scheme in which all matrices are inverted (which increases computing time by five for this trajectory). These solutions are shown by the DASHED lines. In addition the scheme has been compared with the NAG GEAR routine, a highly accurate scheme which uses a backward differentiation formula with a variable step size adjusted to keep the local truncation error below a specified bound. These solutions are shown by DOTTED lines.
The figures demonstrate that IMPACT performs well in comparison with more sophisticated and expensive schemes. Total chlorine and bromine is conserved and only small differences occur between the schemes for individual families or species such as ClOx and BrCl. One problem with using variable step methods in GCM chemistry becomes apparent here. The NAG routine slows down to a standstill after the PSC event on day 5 as it takes smaller and smaller steps. The vertical dotted lines towards the end of day 6 descending from Cly in figure 1 and from Bry in figure 2 show where the job exceeded its 6000 cpu seconds limit on the Cray, after which all the species and families were set to zero. (The IMPACT scheme took 42 cpu seconds.)
Peter Stott & Glenn Carver
We have run the EUGCM from initial data of Nov. 7th 1991 (ECMWF/ISAMS dynamic fields) over the Northern hemisphere winter period. We have then used TOMCAT (which has a much more sophisticated advection scheme than the EUGCM) forced by EUGCM dynamics every 6 hours, to advect CH4, H2O and N2O tracers (initial fields from ISAMS and MLS) on which a parametrized chemistry operates. Domain filling trajectories were also included in the EUGCM run and advected in three dimensions. The combination of trajectory information and tracer fields should allow us to diagnose the motion of long-lived tracers in and around the polar vortex. A comparison of model fields with UKMO assimilated fields and MLS water vapour is being undertaken.
We have also used visualisation techniques in which the evolution of 3-D isosurfaces representing the vortex and the Aleutian high (defined using scaled geopotential height values) can be investigated. EUGCM output and UKMO assimilated data along with vertical slices of tracer values (modelled and MLS water vapour) have been visualised in this way.
Anne Pardaens, Gordon Watson, Peter Stott
We have been given time to use a Cray C94 computer at Cineca, Italy, funded by the European Community. We intend to use our time to make some high resolution runs of the Northern hemisphere winter vortex using the USMM. The number crunching ability of this computer leads us to hope that we may run the USMM at T106 or even T213 resolution. If anybody has any suggestions or comments please get in touch with us as we would like these runs to be of maximum benefit to UGAMP.
Peter Stott, Gordon Watson (pas@uk.ac.ed.met and gcw@uk.ac.ed.met)
Measurements of stratospheric water vapour by several satellite instruments show evidence of an upward propagation of the annually varying tropopause-level saturation mixing ratio (smr). The seasonal-scale phase relationship between water vapour mixing ratio in the lower stratosphere and smr at 100 hPa is well diagnosed by a radiatively based calculation of the transformed Eulerian mean circulation, which confirms the robustness of the calculation and Michael McIntyre's ''tape recorder'' hypothesis that air would be ''marked'' on entry to the stratosphere. Water vapour seems to show that vertical and meridional mixing are small in the tropics, as ''markings'' at the tropopause are retained for up to 18 months. The quasi-biennial oscillation in tropical zonal winds has a small impact on the timing of minima and maxima.
Figure 1 shows time-height sections of 2*CH4+H2O for HALOE and of deviation from time-mean water vapour for MLS and SAGE II. For MLS, the time period spanned is January 1992 to April 1993, and for HALOE, January 1993 to October 1994 using monthly means. The SAGE figure shows two cycles of the annual and semiannual components of the 1986-1991 detrended mean.
MLS water vapour (Fig. 1a) clearly suggests the upward propagation of successive minima and maxima, and as shown elsewhere, the timing of MLS minima and maxima at 46 and 22 hPa agrees well with Lagrangian calculations using the transformed-Eulerian mean (TEM) circulation based on observations. The TEM calculation indicated transit times from the tropopause to 46, 22 and 10 hPa of 8, 14 and 18 months, respectively, plus or minus 1 month.
HALOE (Fig. 1b) not only provides information on the lowest 5 km of the stratosphere, which MLS does not, but also has higher vertical resolution and retrieves methane as well. Fig. 1b confirms the long transit time from 100 hPa to 46 hPa. It also suggests that the drier air entering the stratosphere in northern winter travels upward relatively quickly in its first few months, whereas the somewhat more moist air entering the stratosphere in northern summer travels upward much more slowly.
The image of the ''tape recorder'' in SAGE II (Fig. 1c) is not as clear. Dry air leaves the tropopause in February or March (and moist air leaves in August) and some months later is evident at 19 km. But the only evidence of the ''tape recorder'' in the aerosol layer (20 to 25 km) is the dry anomaly at 22 km in December, which is consistent with MLS, HALOE and the TEM calculation.
An improved version of the two-dimensional isentropic model described by Kinnersley and Harwood [1993] (QJRMS vol. 119, pp. 1167-1193) shows that the QBO can reduce or increase parcel transit times to the middle stratosphere by about a month (Figure 2). Given the zonal mean winds during the UARS period, this reduction or increase partly explains the small discrepancies between the TEM calculation and the timing of minima and maxima in the UARS record.
For a more complete description of this work, ftp cumulus.met.ed.ac.uk, cd pub/ms/LSWV, ls, get all files.
Table of Contents * Previous Chapter * Next Chapter