General Overview
Online Documents
Access to Data
Daily Observation History
Photo Gallery
Related Sites
Rules of the Road
TIDI Web Site
The Final Word
Instrument Design


The distribution of most chemical species in the stratosphere is affected by both dynamical and chemical processes. Conversely, the distribution of certain photochemical species, such as ozone, can influence the radiative budget of the stratosphere, affecting temperatures and motions. Until recently, satellite remote observations of the stratosphere have provided only temperature and constituent measurements. Global horizontal winds have been deduced from temperature fields by using the thermal-wind relationships, which relate the vertical shear of the geostrophic wind components to horizontal temperature gradients. These equations, however, are only a good approximation for large-scale, low-frequency, extratropical flows (D.J. Andrews, Academic Press, 1987). The High Resolution Doppler Imager (HRDI) on the Upper Atmosphere Research Satellite (UARS) is providing the first direct measurement of the global horizontal wind field in the stratosphere, mesosphere, and thermosphere. UARS was launched September 12, 1991, into a 585 km circular orbit inclined 57 deg to the equator (C.A. Reber, EOS 71, 1867, 1990). The spacecraft carries ten instruments designed to study the chemistry and dynamics of the stratosphere and above. Knowledge of the stratospheric wind field will help to determine how transport and mixing influence the ozone budget in the lower stratosphere, and to quantify the mean and eddy circulations associated with the observed semiannual oscillations. In the mesosphere, the wind measurements will contribute to the understanding of the relative role of turbulent diffusion and bulk advection in accounting for mesospheric tracer budgets.

Figure 1
Figure 1. Schematic illustrating the physical processes involved in the HRDI measurement.

The goal of HRDI is to measure the vector winds in the stratosphere (10-40 km), mesosphere and lower thermosphere (~50-120 km) during the day, and the lower thermosphere at night (~95 km) to an accuracy of 5 m/s. The horizontal wind vector is measured by observing the Doppler shift of rotational lines of molecular oxygen along two lines of sight (P.B. Hays, Appl. Opt. 21, 1136, 1982; V.J. Abreu et al., Appl. Opt. 28, 2128, 1989). In addition to winds, temperatures and volume emission rates are obtained in the mesosphere and lower thermosphere, and cloud top heights, effective albedos, aerosol phase functions, and scattering coefficients are determined in the stratosphere. Figure 1 schematically depicts the physical processes involved in the measurement. Depending on the band and atmospheric region, molecular oxygen lines can be observed in either emission or absorption, the latter dominating below ~50 km. Solar radiation incident on the atmosphere is absorbed, exciting molecular oxygen into the equation state. This can occur directly through resonance or photochemical reactions (L. Wallace and D.M. Hunten, J. Geophys. Res. 73, 4813, 1968; A. Bucholtz et al., Planet. Space Sci. 34, 1031, 1986). When this process takes place in the mesosphere or above, where quenching effects are small, the O2 emission lines in the A band of the (equation) transition are observable. The emission in the B (1,0) and gamma (2,0) bands can be ignored because the excited vibrational states are collisionally quenched into the nu'=0 state before the electronic transition takes place (L.C. Lee and T.G. Slanger, J. Chem. Phys. 69, 4053, 1978; M.J. Gauthier and D.R. Snelling, Can. J. of Chem. 52, 4007, 1974; U. Schurath, J. Photochem. 4, 215, 1975). Below ~50 km electronic quenching (equation) rather than emission dominates the loss process. In this case, the absorbed photons are not reemitted, and absorption lines develop in the spectrum. Absorption features in the A (~762 nm, 13,120 cm-1), B (~688 nm, 14,525 cm-1), and gamma (~629 nm, 15,902 cm-1) bands can be observed when solar radiation is elastically scattered by molecules and aerosols into the field of view of the observer. This scattering may occur before or after light is reflected from the surface. The line features generated by the Atmospheric bands (see P.B. Hays, Appl. Opt. 21, 1136, 1982) are generally quite narrow, with a width of ~0.001 nm (2 x 10-2 cm-1), and are therefore suitable for the Doppler shift measurement technique. The HRDI instrument and early results have been discussed elsewhere (P.B. Hays et al., in Digest of the Topical Meeting on Optical Remote Sensing of the Atmosphere, Vol. 4, DC, p. 7; V.J. Abreu et al., Optics & Photon. News 2, 28, 1991; P.B. Hays et al., J. Geophys. Res. 98, 10,713, 1993; P.B. Hays et al., Planet. Space Sci. 40, 1599, 1993; W.R. Skinner et al., Proc. NASA Symp. on Global Wind Measurements, Deepak Press, VA, 1986, p. 129; W.R. Skinner et al., IGARSS '87, 1987; Y.T. Morton et al., Geophys. Res. Lett. 20, 1263, 1993). This paper will review the instrument, discuss some of the important in-flight calibrations, and give examples of the type of data recovery possible. Several atmospheric parameters can be obtained with HRDI, but this paper will deal exclusively with the recovery of winds.

Instrument Requirements

The requirement to measure a Doppler shift on the order of 10-4 cm-1 demands an instrument that has high throughput and high spectral resolution. The Fabry-Perot interferometer that is used in HRDI is such a device (see P. Jacquinot, J. Opt. Soc. of Amer. 44, 761, 1954). High throughput and high resolution are not independent parameters, and optimization must be performed to define the best system. The optimization procedure must consider practical limitations. For example, the resolution (FWHH) of a Fabry-Perot is given by:

equation (1) (1)
and N is a combination of the finesses due to reflectivity, instrument defects and detector broadening. To a good approximation, this is given by
equation (2) (2)
The amount of light collected in the central order (assuming no true absorption) is close to
equation (3)(3)
There are three important features to note from these equations. First, the amount of light collected decreases monotonically as the reflectivity increases. Second, the light in the central order depends only on the free spectral range (which in turn depends on the gap spacing) and the etalon reflectivity, while the resolution depends on the defects as well. Defects increase the interferometer halfwidth and decrease the peak transmission. Third, if the finesses are significantly different, then the resolution is determined by the smallest finesse. A practical rule is to try to make the finesses approximately equal. The HRDI detector is a multi-channel device which has 32 channels equally spaced in wavelength. The number of channels was limited by telemetry and manufacturing capabilities. The aperture finesse is approximately the number of channels per order. This was set to be 30 at 762 nm (i.e., ~1 order on the detector) and it decreased linearly with wavelength to 25 at 630 nm. The Fabry-Perot plates were manufactured to a matched flatness of ~lambda/200, and in order to ensure that the etalons would survive launch stresses, they were firmly attached to their mounts. This had the unfortunate effect of significantly stressing and distorting the plates so that the defect finesses were about 20. The reflectivity of the etalons was about 0.90 at all wavelengths of interest, corresponding to a reflectivity finesse of about 30. These combine to give an overall finesse of about 12. The large degree of distortion provided by the etalon mountings was unanticipated and the etalons were coated to a higher than optimal value.

The quantity of light collected by a Fabry-Perot is proportional to the effective area of the etalon and the solid angle passing through it. The solid angle through the etalon is

equation (4)(4)
HRDI examines a spectral range of 0.5 cm-1 at ~13,100 cm-1. The solid angle is then approximately 2.4 x 10-4 sr and, with an effective etalon diameter of 8.9 cm, the area-solid angle product is 1.49 x 10-2 cm2 sr. This number must be matched by all other optical elements in the system. This requirement is readily met by the rest of the interferometer optics; the only real concern is the telescope that observes the atmosphere. The diameter of the telescope is limited in size by mass constraints, cost of the primary mirror, and the length of the baffle. A large-diameter telescope requires a long baffle adding its own mass and moment of inertia, which is a consideration since the telescope must move rather rapidly and with a minimum of wobble. Since HRDI is a limb viewer, the vertical field of view is limited by the vertical scale of the features in the atmosphere which the instrument is designed to study. The scale height of the atmosphere in the stratosphere, mesosphere and lower thermosphere ranges from 5 to 8 km. A vertical field of view of 0.12 degrees corresponds to a vertical range on the limb of 5 km at a satellite altitude of 600 km. The horizontal field of view can be much wider, since typical horizontal scales in the atmosphere are much larger than vertical features. Since HRDI measures the winds by determining the Doppler shift, care must be taken to properly account for the Doppler shift provided by the spacecraft motion. If the field of view is too large there will be a significant Doppler shift variation across the field of view. A consideration of these factors leads to a telescope design which incorporates a primary 17.78 cm in diameter and field of view of 0.12 x 1.3 degrees (a secondary field of view of 0.12 x 0.12 degrees is available for star observations). The telescope is an off-axis Gregorian which minimizes small-angle scattering, and the area-solid angle product is then 1.1 x 10-2 cm2 sr. This demonstrates that the etalon and telescope are not optimally matched. The entire etalon (area and solid angle) is filled by the use of a fiber optic bundle which illuminates it with a lower intrinsic brightness than if the etalon and telescope were matched.

Determination of Wind Shift

In order to determine winds, which is the primary product of the HRDI instrument, it is necessary to determine the Doppler shift of an absorption or emission line. Figure 5(a) shows the signal observed by an emission line on the IPD with and without a Doppler shift. The unshifted line is centered on channel 16 and the shifted line is moved by 20 m/s. Figure 5(b) clearly demonstrates that the shift has a very small effect at line center and in the wings, and is most notable on the sides of the line. The Doppler shift causes the signal in some channels to increase and to decrease in others. The proper interpretation of these signal changes allows the Doppler shift to be determined. The algorithm used to recover the Doppler shift is complex, and only a conceptual demonstration of the technique is presented here. The signal in the presence of a shifted emission or absorption line is given by

equation (5) (5)
Equation 5 can be inverted in a least square fashion using all channels and converted by the Doppler equation to give
equation (6)(6)
The statistical uncertainty of velocity can be estimated by assuming the only source of error is photon statistics in Ci(U). This is the largest source of error and can adequately describe the statistical error. Using Poisson statistics which describe photon noise, the uncertainty in U is given by
equation (7) (7)
Equation (7) demonstrates that to minimize wind errors the signal should be as bright as possible and the slope of the signal should be large. The instrumental width (~0.001 nm, 0.02 cm-1) is significantly greater than the shift due to atmospheric winds. It is important to recognize that HRDI measures Doppler shifts on the order of 10-5 nm, and is not attempting to resolve spectral features of this magnitude. Since the width of the spectral line depends on convolution of the instrument width and the atmospheric line shape, there is no reason to make the instrument resolution much narrower than the line as the resultant signal will not become sharper and the signal-to-noise ratio will decrease.

Fig. 5. Illustration of the effect Doppler shifts have on the spectrum observed by HRDI. a) Spectra with and without a small Doppler shift. b) The difference between the two spectra. The center and wings are unaffected, but the sides show a dramatic change.

HRDI Home Page
Last Update: Friday, 24-Oct-2003
Page prepared by: meyerkm@umich.edu