From CedarWiki
Jump to: navigation, search

Davis, Antarctica Spectrometer


Contact Persons

Web Pages


The Davis Czerny-Turner scanning spectrophotometer is operated by members of the Australian National Antarctic Research Expedition and supported by the Antarctic Science Advisory Committee and the Australian Antarctic Division.

Instrument/Model Description

The Davis, Antarctica (68.48S, 77.97E; 25 m alt) Czerny-Turner (CZT) scanning spectrophotometer obtains mesosphere temperatures from ratios of the hydroxyl (OH) (6-2) band during nighttime hours. The spectrophotometer was in campaign operation March to October 1990 and April to August 1994, and has had continuous winter operations since March 1995. (See Greet et al., 1998; French et al., 2000; Burns et al., 2002.)

Hydroxyl (6-2) band rotational temperatures are derived using Langhoff et al. (1986) transition probabilities, which are approximately 2 K higher than the temperatures derived by French et al. (2000) using experimentally derived P1 branch ratios with a slit width of 100 microns. The hydroxyl layer is centered at a mean height of 87+/-4 km with a mean thickness of 8 km. WINDII data showed that the hydroxyl layer often (up to 25% of the the time) has 2 peaks between 83 and 93 km (She and Lowe, 1998). The brightness of this layer usually decreases during the night.

On day 359 of 2001 at 87 km altitude, the apex magnetic coordinates were (-64.1, 152.4) degrees. The magnetic inclination and declination angles were -81.7 deg and 21.8. deg. The magnetic local time at 0000 Universal Time (UT) is about 0504 MLT. The solar local time (SLT) is UT plus 5 hours and 12 minutes (77.97/15.=5.198).

The spectrophotometer has a 6 degree field-of-view (fov) that was aligned 30-degrees above the horizon in a direction +130 E from Davis away from auroral precipitation in 1990 (Greet et al., 1998). An order separating filter was not used in 1990. Starting in 1994, the optical axis was aligned in the zenith, with a resultant area of about 8 km x 8 km at 87 km. In practice the field of view is defined by the diffraction grating. The diffraction grating rotates about the vertical axis as it scans, so that the area of the grating directly viewing the sky decreases with wavelength (when the grating is at vertical the fov is zero). Thus the fov corresponding to the grating vertical axis varies with grating rotation angle and consequently with wavelength. The instrument is aligned towards the SE (+130 E), but rotates about the vertical as the spectrum is sampled, so these observations are considered to be in the vertical.

A cooled (-28C) GaAs photomultiplier tube is used for photon detection. Photons are usually counted for dwell times of 0.1 s using a slit separation of 250 microns. In 1990, 5 sequential scans from 837.5-856.0 nm at 0.005 nm steps were accumulated in about 54 min. In 1994, the scans were cut off at 851.5 nm reducing the acquire time to about 48 min in 1994 and 38 min in 1995-1996. Starting on day 269 (Sep 25) of 1996, a single piece-wise scan of small spectral regions including the P1(2) and P1(4) lines (~839.9 nm and ~846.5 nm) and selected backgrounds in steps of 0.001 nm for ~6 min was begun with a 12 sec dark spectral time from 839.0-839.1 nm inbetween scans. Starting day 307 (Nov 02) of 1996, the P1(5) line (~850.5nm) was included, although the scan time was still ~6 min. The scan was lengthened to ~7 min or more starting 1 July 1997 when the slit width was changed to 100 microns until 17 Aug 1997. However, the scan remained about 7 min thereafter. Two scans are always used, where most scans are used twice for integration times of about 14 min every 7 min.

The scan segments for 1997 were:

  • File Prefix Feature Start Stop Step Dwell Period TotalTime
  • E E E sec sec sec
  • Scan Structure
  • DRK Dark Scan 8390 8391 0.1 1 1.08 11.88
  • P12 P12 line 8396.5 8401.5 0.1 1 1.08 55.08
  • BP12 P12 background 8405.4 8409.4 0.1 0.5 0.58 23.78
  • BP14 P14 Background 8437.5 8441.5 0.1 0.5 0.58 23.78
  • OI Auroral Line 8443.8 8448.8 0.1 0.5 0.58 29.58
  • P14 P14 Line 8462.5 8468 0.1 1 1.08 60.48
  • BP15 P15 background 8477.5 8487 0.1 0.5 0.58 55.68
  • FRA Fraunhoffer sample 8494.5 8501.5 0.1 0.5 0.57 40.47
  • P15 P15 Line 8502.1 8507.7 0.1 1 1.08 61.56
  • The total time is 362.29 sec or 6 min and 2 sec, so the integration time for 1997 was estimated to be 2*6min*60sec/min = 720 sec. Starting after 10 UT on 1 July, 1997, the average time inbetween scans was longer, 7 min or more. Before 2001, the time inbetween scans could be a couple of minutes or more while the data was stored on a directory before archival. Starting in 2001, all scans were stored and archived immediately.

The midpoint UT time for 1990-1996 (up to day 268) data is the time of the OH(6-2) P1(2) emission line near the beginning of the middle scan. From late 1996, the midpoint UT is the file close time of the (12 second) dark sample file inbetween the two spectra used. The begin and end times are estimated using the average scan times, even if the integration times are longer with storage times inbetween scans.

The relative brightness or intensity (code 2505 for kindat=17002, or code 2507 for kindat=17001 in 1990) is the sum in counts of the P1(2), P1(4) and P1(5) lines measured with the background subtracted.

The spectral response was determined starting in 1996 by scanning a low brightness source (LBS) which is calibrated by the Australian National Measurement Laboratories (NML) each summer. Final response correction curves for a given year are determined from consideration of the NML calibrations at the start and end of that year. Initial temperature results are available using the previous years calibration curve. Data from the present season are thus available from the contact persons, but are not in the CEDAR Database since they will be revised after the calibrations are complete at the end of the season. The data for 1990, 1994-1996 will be reanalyzed following the review of the instrument response corrections, which should only change the temperatures slightly.

The inter-year accuracy is better than 1 K for spectral response determinations since 1996, with accuracies of about 2 K for 1990 and 1994-1995 using a quartz-halogen lamp calibrated in 1996. The instrument function was defined at 632.82 nm using a frequency stabilized laser. This was scaled to the OH(6-2) P1(4) wavelength, where the full-width-at-half-maximum (fwhm) of the instrument function was determined to be 0.155 nm using a 250 micron slit separation (French et al., 2000).

Temperatures are derived as a weighted average of rotational temperatures from the 3 ratios of P1(2)/P1(4), P1(2)/P1(5) and P1(4)/P1(5). Only the P1(2)/P1(4) lines were used between days 269 and 306 in 1996. The P1(3) line is not used because of contamination with the non-thermalized OH(5-1) P1(12) emission (see Greet et al, 1998). Since there is an average decrease in intensity during the night, the intensities are linearly interpolated to a common time between successive ~7 min scans after 1997. The average decrease in hydroxyl intensity across selected spectra implied that the 1990 operating mode underestimated the temperatures by about 1 K. However, interpolation between scans versus summing 5 scans in 1996 resulted in temperatures about 0.7 K lower, implying a small increase in average hydroxyl intensities during spectra acquisition instead of a decrease. The weighted temperature (code 812) is given along with a weighted standard deviation (code -812) from the 3 ratios which is less than 15 K, and a weighted counting error (code 4131) which is less than 10 K for analyzed scans (kindat=17001 for 1990 and kindat=17002).

Since the temperature is calculated from the ratio of two emission lines via equation 1 in French et al (2000, Annales Geophys 18,1293-1303), we can calculate an equivalent temperature error, due to the uncertainty in the estimate of the emission line ratio .. ie by substituting Ia/Ib in that equ with Ia/Ib 1 (Ia/Ib)*sqrt(REa^2 + REb^2) where I are the intensities and RE are the relative counting errors for each line. So that gives you a counting error for each set of the three line ratios measures P1(2)/P1(4), P1(2)/P1(5) and P1(4)/P1(5). Each spectrum determines a weighted mean temperature using the counting errors of each of the three ratios. On average the weighting factors are 0.32 for P1(2)/P1(4), 0.57 for P1(2)/P1(5) and 0.12 for P1(4)/ A weighted counting error is derived as sqrt(3/2)*sqrt(1/sum(weighting factors)). P1(5). The 3/2 is there since there are 2 independent temperature estimates out of the 3 ratios.

Scans are accepted under the following selection criteria:

    1. They have sufficient photon counts for the statistical uncertainty in the temperature (code -812) to be > 15 K. Insufficient photon counts usually indicate cloud cover.
    2. The 3 line-ratio temperatures must be within 10 K
    3. Limitations on the slope and absolute magnitude of the backgrounds High background levels indicate significant aurora or scattered moonlight
    4. For piece-wise spectra: Limitations on rates of change of line and background intensities and derived temperatures

Cloud conditions usually give warmer temperatures, but the average yearly temperature difference is less than 0.3 K. Similarly, auroral contamination can give at most a 0.5 K temperature increase. The cloud conditions (0-5) are in the Davis quality code 4132:

    1. 0 = clear
    2. 1 = haze or thin cirrus (through which stars can be seen)
    3. 2 = patchy or intermittent cloud
    4. 3 = overcast
    5. 4 = snow
    6. 5 = unknown (goes to -32767 or missing)

The conditions are determined through visual observation and reference to an all-sky video systm.

The nightly average temperatures (code 812 in kindat=17011) are a simple mean of the nightly observations, where the minimum number of points per night is 3 for 1990 and 1994, 5 for 1995-6, and 10 from 1997 and later. The error bar (code -812) is the unweighted standard deviation (SD), so
SD = sqrt [ sum of (T-Tmean)^2 / N ]
where N is the number of measurements (code 415) used to derived the nightly average in code 812. Code 4133 is the Standard Error in the Mean (SEM), where SEM = SD/sqrt(N) (or code 4133 = code -812 / sqrt(code 415). The UT for each average (code 34) is the average of the UTs for the observations. Users are free to calculate their own weighted average nightly temperatures, possibly using the maximum of the individual SDs or counting errors. The integration time for these averages (code 61) is less than or equal to the time over which there are good observations (code 4134) in the night.

There is discussion in the literature at present about which transition probabilities it is appropriate to use. We have adopted an approach deriving temperature as often as possible (moon-up, minor auroral contamination, all cloud conditions where the selection criteria are satisfied; see Burns et al., 2002). Our analysis techniques are continually being altered and hopefully improved.

Data gaps:

    1. Between 27 May (DOY 147) and 2 Aug (DOY 214) 1995 the instrument function of the instrument was degraded because it was bumped out of alignment resulting in a wider instrument function than normal.
    2. Between 27 Jun (DOY 179) and 19 Jul (DOY 201) 1996 extensive calibrations were undertaken on the instrument and no spectra were collected.
    3. From 25 Sep (DOY 269) to 1 Nov (DOY 306) 1996 piece-wise scans but incorporating only P1(2) and P1(4) lines were undertaken.
    4. Between 1 Jul (DOY 182) to 17 Aug (DOY 229) 1997 the slit width was reduced from its normal 250 microns to 100 microns for a specific research program (French et al., 2000).


Burns, G.B., French, W.J.R., Greet, F.A., Phillips, F.A., Williams, P.F.B., Finlayson, K. and Klich, G, Seasonal variations and inter-year trends in 7 years of hydroxyl airglow rotational temperatures at Davis station (69S, 78E), Antarctica. Journal of Atmospheric and Solar-Terrestrial Physics, 64, 1167-1174, 2002.
French, W.J.R and Burns, G.B., The influence of planetary scale variability on the long-term trend assessment of hydroxyl temperatures over Davis, Antarctica, Journal of Atmospheric and Solar-Terrestrial Physics, 66, 493-506, 2004.
French, W.J.R., Burns, G.B., Finlayson, K., Greet, P.A., Lowe, R.P. and Williams, P.F.B, Hydroxyl (6-2) airglow emission intensity ratios for rotational temperature determination. Annales Geophysicae, 18, 1293-1303, 2000.
Greet, P.A., French, W.J.R., Burns, G.B., Williams, P.F.B., Lowe, R.P. and Finlayson, K, OH(6-2) spectra and rotational temperature measurements at Davis, Antarctica. Annales Geophysicae 16, 77-89, 1998.
Langhoff, S.R., Werner, H-J, Rosmus, P, Theoretical transition probabilities for the OH meinel system. Journal of Molecular Spectroscopy, 118, 507-529, 1986.
She, C.Y. and R.P. Lowe, Seasonal temperature variations in the mesopause region at mid-latitude: Comparison of lidar and hydroxyl rotational temperatures using WINDII/UARS OH height profiles. Journal of Atmospheric and Solar-Terrestrial Physics, 60, 1573-1583, 1998.

Summary Plots for Davis, Antarctica Spectrometer

Brightness in counts is also available, but not plotted.

-Revised 04 Sep 2004 by Barbara Emery