|
||||||
|
|
||||||
|
||||||
|
|
||||||
The pressure sensor used in the buoy was a quartz beam which oscillates at a resonant frequency dependent primarily on the pressure, and to a lesser extent, on the temperature. These sensors were selected because of their good long term stability and low power consumption. For properly aged sensors the advertised long term stability is on the order of 0.1 mb per year, (Paros 1976). Several attempts were made to evaluate the field performance of the sensors. Before describing the results of these attempts a word about the original calibration procedure is required.
The sensors were calibrated by the manufacturer (Paroscientific, Redmond, Washington) by measuring the frequency of vibration of the quartz beam at pressures 900 mb, 950 mb, 1000 mb, and 1050 mb and at temperatures -50-,-30-, -15-, 0- and 21 degrees Celsius. These measurements were made in the fall of 1978, and were supplied to the buoy manufacturer (Polar Research Laboratory, Santa Barbara, CA) with the sensors. In the buoy, the pressure measurement was made by counting the sensor frequency for a time interval determined by a fixed number of cycles of the buoy clock. Since this clock was also temperature dependent, it was calibrated by the buoy manufacturer at the same set of temperatures. Thus the measured number of cycles n at pressure p and temperature T was
n = g(T)*f(p,T)
where g was the time interval and f was the sensor output frequency.
The temperature was determined using a thermistor inside the housing of the pressure sensor, the output m of which was related to the temperature by
m = h(T).
Thus, given the measurements m and n and the calibration data f, h, and g, supplied by the manufacturers, it was possible to invert these equations to find the pressure and temperature. This was done by the buoy manufacturer and resulted in a polynomial expression which fit the original pressure data to within about 0.2 mb.
As a check of this procedure the manufacturer made readings from the completed buoys, calculated the pressures and compared these against a reference mercurial barometer. In most cases the results agreed to within 0.5 mb. Where larger discrepancies occurred, a correction was applied to the constant term in the polynomial. These polynomials were supplied to Service Argos and were used by them to convert the raw data into physical units.
Most of the buoys were tested again in Fairbanks, Alaska immediately prior to deployment. Results from these tests indicate discrepancies between the buoys of up to 2.5 mb. These errors must either be attributed to errors in the original calibration data represented by g, f, and h or to actual changes in the characteristics of the sensors or buoy. Although the first possibility cannot be ruled out, a review of the original pressure calibration measurements embodied in f only turned up one error (for buoy 1913) and it had only a minor effect. A thorough test of the second possibility would have required repeated visits to each of the buoys, and this was not possible for logistic reasons. Several isolated pieces of evidence are available however.
The only long term comparison was for the reference buoy at Barrow. If the reported pressures from the buoy are interpolated to the synoptic time intervals, they can be compared to the readings reported for the Barrow weather station. The weather service readings were made from an aneroid barometer, checked periodically against a mercurial barometer, and adjusted to sea level. The reference buoy was located on top of the weather station at an elevation of about 9 m above sea level. Generally buoy readings were not adjusted to sea level, but for the sake of this comparison 1.0 mb was added to the reference buoy readings. When this adjustment was made the pressure indicated by the buoy during the spring of 1979 was 0.8 mb lower than the pressure reported by the weather service. A random variation of 1 standard deviation = 0.15 mb was superimposed on this steady bias of -0.8 mb. The precision of the readings (1 count = 0.15 mb) accounts for the random fluctuations. A similar comparison in the fall of 1979 gave a bias of -1.0 mb and 1 standard deviation = 0.15 mb. There is evidence then that this sensor has drifted -.2 mb over a seven month period.
On three occasions it was possible to visit buoys after they were deployed. On 29 April 1979, a flight was made from the drifting station Fram I to the vicinity of buoy 1916. The pressure reported by the buoy was 0.7 mb lower than the value read on the portable reference barometer. This was not consistent with the apparently high reading of this buoy at Fairbanks.
On 25 May 1979, buoy 1903 was only 20 km from the drifting station LOREX - Snowsnake. The pressures reported by the buoy and by the observers at Snowsnake agreed to within 0.1 mb. This was consistent with the negligible offset observed for this buoy at Fairbanks.
On 11 December 1979 a flight was made to the vicinity of buoy 1908. The results show this buoy reading about 1.5 mb low, and this was consistent with measurements made at Fairbanks ten weeks earlier.
It was possible to monitor the behavior of the buoy clock -- which figures into the calibration through the function g(T) -- without visiting the buoys. This was because the buoy transmission times were determined by this clock. A change in the clock frequency results in a change in the time interval between transmissions. The times of reception were recorded by the satellite and reported with the data from Service Argos with a resolution of one second. Since the temperature is also known, it is possible to test the function g(T). The results of tests performed on most buoys indicated that the buoy clock varied with temperature in a manner which was consistent with its original calibration.
In summary, the evidence which bears on the performance of the pressure sensors is not entirely consistent. Nevertheless, the bulk of the evidence suggests that the sensors sustained offsets of typically 1 mb subsequent to their initial calibration but prior to deployment and that these offsets remained constant thereafter. The corrections in Table I have been applied to the data based on this interpretation. Since the level of confidence in these corrections is low, the data must be used with the understanding that unknown biases of about 1 mb are probable.
Buoy ID: 1901 1906 1907 1908 1909 1913 1914 1915 1916 1918 1919 1926 Pressure Correction mb 0.8 -1.1 1.5 -0.6 -1.5 -0.4 0.3 -0.5 -1.2 -0.8 0.8 Temper- ature Correction degrees C +6.4 No corrections were applied to buoys 1902, 1903, 1905, 1911, 1917, 1920, 1923, 1924, 1925 or 1927.
