Project Website http://www.haystack.mit.edu/obs/optics/index.html
The Millstone Hill Fabry Perot interferometer was operated by MIT in cooperation with the University of Pittsburgh through 2005. The interferometer is located near the Millstone Hill incoherent scatter radar at latitude 42 degrees 37 minutes North (42.62) and longitude 71 degrees 27 minutes West (-71.45). Mean local solar time lags UT by 4 hour 46 minutes. The local magnetic field has a 15 degree variation to the West and an inclination of 72 degrees.
The FPI is a 100 mm aperture pressure (index of refraction) tuned interferometer. The pressure tuning is done by moving a piston. Both upward and downward pressure scans are used. A multi-aperture exit plate selects 5 interference orders for field widening. A cooled gallium-arsenide photocathode photomultiplier is the detector. The FPI observes the sky through a two-mirror pointing head. A frequency stabilized laser is used to monitor the stability of the FPI and to determine the instrumental function for derivation of temperatures. The FPI is operated by a PC, which controls the mechanical drivers for the instrument, stores the data and performs some pre-processing. A queue file contains a list of commands to the FPI which are used to run the instrument on specific days in special modes. Parameter files can be used to specify observation sequences. The FPI is usually operated every night and data from clear nights are extracted and analyzed. The analyzed data are sent to the CEDAR database The instrument was built at the University of Pittsburgh and has been described in the literature.
Spatial scanning of the interferometer is performed by a two mirror pointing head which allows us to select any point in the sky for observation. The standard observing mode uses a vertical measurement and 4 measurements at an elevation of 30 degrees from the horizon looking at 45 degrees from the cardinal points (azimuths 45, 135, 225 and 315). This sequence provides 2 orthogonal vector observations at each of two different latitudes. This sequence is marked on data sets as "using pointing head sequence 45DEGREE". Several variations of this sequence exist. For more information check with the data analyst.
Data are collected on both the upward and downward ramps. The data are added synchronously until either (1) the required signal is obtained, (2) a large background is observed or (3) a time limit (typically about 10 minutes) is reached. To minimize pressure sensor hysteresis effects, the upward and downward scans are summed separately. A frequency stabilized HeNe laser (632.8165 nm) is used to measure the instrumental response and to monitor instrumental drifts.
Analysis of the data is a three step process. First, all the data from the frequency stabilized laser are fit to a parameterized Airy function, producing a table of the instrumental parameters as a function of time throughout the night. Second, a parabolic numerical least squares fitting process is then used on the nightglow data, based on the interpolated instrumental parameters. This method gives 4 parameters: a doppler shift of the nightglow from the shift in the measured peak, a relative intensity of the nightglow from the signal integrated over the peak, an effective temperature of the neutral atmosphere from the doppler width of the measured spectrum and a continuum background signal from the baseline of the profile. In the third analysis step, the doppler shift of the nightglow line is interpreted in terms of a wind.
Log10 relative emission intensity (parameter 2506 in the CEDAR database) is the integration of the fitted line profile over the free spectral range of the instrument. This is only a relative intensity parameter, intended for comparison of intensities during a single night, or perhaps over periods of a week or two. Changes or drifts in sensitivity are not removed from this number, so comparisons between different nights are not advised. An order of magnitude estimate of the calibration is 10 of these units per rayleigh, however, this is only a very rough approximation.
Winds are calculated by measuring the difference in line position between the fitted line and a zero velocity reference. The zero velocity reference is generated by taking the positions of all the fitted lines from zenith observations and smoothing and interpolating them as a function of time. This assumes that the vertical velocity is small compared to the resolution of the interferometer. For nights in which the quality of the vertical measurements is poor, or in which there are not enough vertical measurements for a good smoothed reference, the vertical measurements may be supplemented by an average of measurements in opposite directions. This assumes that the wind field is uniform over the observation points, i.e. without divergences. The method used to obtain the vertical reference is flagged in the CEDAR database by KINDAT. Further, KINDAT is used to differentiate between "direct" and "derived" data (see below). KINDAT = 7001or 17001 velocity reference uses only vertical measurements, while KINDAT = 7002 or 17002 velocity reference uses combined measurements. KINDAT = 7001 or 7002 is used for "direct" data and KINDAT = 17001 or 17002 is used for "derived" data.
Since data from both the upward and downward pressure ramps are analyzed separately, there are generally 2 wind and temperature measurements at each time, which are combined to give a single wind and temperature. The final parameter is the geometric mean of the original 2 points. The uncertainties could be reduced by a factor of the square root of 2 if the individual uncertainties were the same. Another measure of the uncertainty is the difference between upward and downward ramps. This is combined with the individual statistical uncertainties by taking half of the square root of [(the sum of the squares of the uncertainties) plus (the square of the difference)]. The resulting value is presented in the CEDAR database as the uncertainty.
The smoothing and interpolation algorithm used in the analysis is a Hanning filter. This filter has been described in the literature (Sipler et al., op. cit. 1991).
Data in the CEDAR database are presented in two records: "direct" and "derived" data. Geographic winds taken from observations in the cardinal directions are considered direct measurements. Two or more measurements may be combined to obtain "derived" data, e.g. one direct meridional measurement and a smoothed interpolated zonal measurement may be used to calculate a wind vector, from which a geomagnetic meridional wind is determined. A similar method is used to calculate geomagnetic zonal winds. The geomagnetic winds computed in this way are considered derived data. Similarly, when data are taken at 45 degrees to the cardinal points, two measurements from orthogonal directions are used to define a vector, from which both geographic and geomagnetic winds are determined. Again, these are derived data.
When data are taken in the cardinal directions, the winds derived from the measurements are either geographic meridional or zonal. In order to determine the magnetic meridional or zonal wind, both components of the wind are needed. In the case of a meridional measurement, the zonal winds through the night are smoothed and the resulting curve is used to interpolate the zonal wind at the time of the meridional measurement. Similarly, the meridional winds are smoothed and interpolated to obtain magnetic zonal winds. The smoothing and interpolation algorithm is the Hanning filter. The smoothed wind is not presented in the database as derived data. These points are listed as missing data. If interpolation is necessary, check with the data analyst.
Latitudinal gradients are seen frequently in the winds. When this occurs and data are being collected at 45 degrees to the cardinal points, the NW and SW data points and the NE and SE data points, which could be used to resolve a wind vector, are not used since they represent different latitudes and therefore different winds. The NW-NE points are from the same latitude and therefore the gradient should not affect the wind determination from these points. Similarly, the SW-SE points are used. When such gradients are observed, a note will be placed in the catalog record for that data set. If no such gradients are observed, it is possible to use any orthogonal pair of measurements to determine the wind vector.
The beginning azimuth (CEDAR database parameter 132) and ending azimuth (parameter 133) given for the derived data represent the two azimuths used in the derivation. The elevation (parameter 140) is the average of the two elevations used in the derivation, however, none of the data are combined for a derived wind if the elevation of the two points differs by more than 5 degrees. Since these data use more than one observation point, the calculation of winds assumes spatial uniformity in the wind field, at least over the points used in the derivation.
Uncertainties in the derived parameters are purely statistical, and do not reflect possible systematic errors. The wind uncertainty is calculated from the data by considering the accuracy of determination of the center of gravity of a line (Hernandez, "Fabry-Perot Interferometers", Cambridge University Press, 1986). Temperatures and temperature errors are estimated by taking a fourier cosine transform of the data and fitting the logarithm of the coefficients to a straight line (as described in Hernandez, op. cit.) The slope of the line gives a temperature estimate and the error in the slope gives the uncertainty in the temperature. The temperature derived from the slope is used as a first estimate in the parabolic least squares fitting process. If the temperature determined from the slope is outside of reasonable limits, the fitting process starts at 1000K. If the temperature uncertainty is greater than 200K, the temperature is listed as missing data. Such temperatures are typically from very noisy data and are considered unreliable.
A tilting filter photometer is also operated in conjunction with this system. The photometer is boresighted with the FPI to give a record of intensities through the night. The photometer data can be used to derive both nightglow and background intensities. This instrument is primarily useful in defining the clear or cloudy periods of any given night. Data from this instrument are not sent to the CEDAR database, but may be available from the Millstone Hill Observatory.