Data Analysis Method

The analysis of the COSMIC data presented below is broadly similar to that used with the previous generation CHAMP satellite (Hei et al. 2008), except that a smaller grid cell size and higher temporal resolution are employed. The raw profiles are interpolated to 1 km vertical resolution both for computational reasons and because the original ~50 m resolution data are not height independent (Baumgaertner and McDonald 2007). The data are binned into grid cells of size 20° x 5° x 7 days, from which the mean background temperature T is calculated. The temperature perturbation of an individual profile T' is calculated in the following way. A profile's T is subtracted from T and the linear mean is removed using a least squares fitting method. Any missing data points are linearly interpolated across (in reality this is not often necessary to do). The data are then Welch windowed to reduce spectral leakage before being 7 km vertically high-pass filtered. The resultant perturbation profile is referred to as T'. These perturbations from the mean are interpreted as being due mainly to gravity waves. The potential energy per unit mass Ep is then calculated using:

where N is the Brunt-Vaisala frequency and g is the gravitational acceleration. Ep is calculated over 7 km vertically and then stepped up by 1 km, thus consecutive heights are not independent. So Ep can be referred to as the mean specific (per unit mass) potential energy. This analysis procedure is stepped forward in time by one day, therefore consecutive days are also not independent. This time resolution compares favorably with that used with GPS/MET (seasonal) and CHAMP (monthly).

T' may be obtained by removing an individual low pass filtered Tprofile and then forming the relevant grid cell perturbation average (Tsuda et al. 2000; de la Torre et al. 2006). Polynomial fitting individual T profiles to obtain T', which are then averaged over the appropriate grid domain result in Ep differences of < 0.4 J kg-1 except around the tropopause, when compared with forming the grid cell mean T firstly. Seven day averaging is optimum given the data constraints (these results are not shown).

Two examples of COSMIC profiles and their resultant Ep are shown in Fig. 1 during August 2007. The top row shows data obtained at 50°S while the bottom row a)3

I 25

30

30

y

25

c

25

/

25

<

20

<

20

20

i

15

^

15

15

190 200 210 220230240

190 200 210 220230240

2 3 4 5 6 7 xlO"4 rod's"'

Fig. 1 (Top row) Profile obtained in the grid cell [260°E, 50°S] at 0633 UT, August 25, 2007. (Bottom row) Profile obtained in the grid cell [160°E, 5°S] at 0103 UT, August 28, 2007. From left to right, the panels show raw T and background T, filtered T', N2, and Ep

shows data at 5°S. Differences between the individual profile's Tand the 20° x 5° x 7 day background T are readily apparent in Fig. 1a,e. Note also the sharp cold-point tropopause at 16 km in the tropical profile of Fig. 1e. The resultant T' perturbation profiles are shown in Fig. 1b,f respectively, which together with the N2 of Fig. 1c,g are used to calculate Ep. The Ep at 50°S in Fig. 1d shows a maximum of 2.2 J kg-1 at 32 km. This large Ep probably is a result of gravity wave emission, Doppler shifting, and reduced critical level filtering activity associated with the stratospheric polar night jet (Baumgaertner and McDonald 2007; Hei et al. 2008). In the tropics (Fig. 1h), the Ep shows a maximum around 16 km as a result of the cold-point tropopause and the low N2. Another maximum of 3.5 J kg-1 is observed at 33 km. This is due to critical level wave interactions which depends upon the phase of the Quasi-Biennial Oscillation (QBO), to be discussed below.

NCEP re-analysis zonal wind velocity u data from 1000 hPa to 10 hPa (about 31 km) (Kalnay et al. 1996) are used for comparison with the COSMIC Ep. Use is also made of Outgoing Longwave Radiation (OLR) data in the tropics, which is widely used as a proxy for deep convective activity (Wheeler and Kiladis 1999).

0 0

Post a comment