1. Introduction
Surface pressure is one of the few meteorological variables not monitored from space, despite calls by the World Meteorological Organization (WMO) for global measurements at 50-km spatial resolution with 0.05-kPa accuracy (WMO 1993). The situation is particularly acute in the Southern Hemisphere, where large expanses of ocean have relatively few surface observations from floating buoys and merchant ships. The principal reason why pressure has not been measured from space is accuracy; if they are to be useful, pressure fields must be measured with accuracy 0.1 kPa in a typical value of 100 kPa. Such accuracy is daunting for remote sensing techniques. Consequently, greater effort has been invested in measuring the gradient of surface pressure, via surface winds, than the pressure itself, and microwave scatterometers now give wind speed with accuracy of approximately 2 m s−1 that can be used to interpolate between pressure measurements from floating buoys.
Nevertheless, several attempts have been made to demonstrate the feasibility of measuring the pressure field from space. Methods include passive microwave sensing of the O2 mass in a column using the 60-GHz O2 absorption line (Peckham et al. 1983), an active lidar system tuned to the O2 A band (Korb et al. 1989), limb sounding in CO2 channels (Allam and Houghton 1988), and passive remote sensing using the O2 A band (Mitchell and O’Brien 1987; Bréon and Bouffiès 1996). This paper focuses on the O2 A band (759–771 nm), lying at the boundary of the far red and near infrared, and presents the results of airborne experiments designed to test the predictions of the modeling by Mitchell and O’Brien (1987). High-resolution A-band spectra of sunlight reflected from the sea were obtained with a grating spectrograph directed toward sunglint from a research airplane. It will be shown that the A-band spectra allow the pressure altitude of the airplane, used here as a proxy for surface pressure, to be tracked with precision of approximately 0.1 kPa under a wide range of surface and atmospheric conditions.
Measurement of surface pressure from reflected sunlight is difficult because radiance reflected to space has two sources: reflection from the surface and scattering by the atmosphere. The scattered component generally traverses a shorter optical path and causes an underestimate of the surface pressure. Mitchell and O’Brien (1987) argued that the two sources might be distinguished with high spectral resolution observations, based on the following principle. When observations are made at frequencies where the absorption optical thickness of the atmosphere is large, radiance reflected from the surface is highly attenuated by absorption along the ray path, leaving radiance scattered by the atmosphere as the dominant contribution to the radiance measured from space. Conversely, at frequencies where absorption is small, the dominant component of the radiance comes from reflection at the surface. Consequently, the breakdown of radiance into its reflected and scattered components can be effected from a combination of observations at increasing absorption optical thickness.
In practice, the prospect of pressure measurements from space with an accuracy of 0.1% is challenging for several reasons.
Every scene would have to be checked rigorously for cloud—a difficult task in itself. To this end, O2 A-band data would have to be supported by data with high spatial resolution (but low radiometric precision) from a bore-sighted space camera.
Equally important are the scattering properties of aerosol, often poorly known and difficult to observe. In principle, the vertical distribution of the extinction coefficient, single scattering albedo, and phase function should be known.
High precision in pressure is not possible without high radiometric precision. However, the latter requires long integration times and large footprints within which the natural variability of reflectance is potentially another source of noise. Consequently, the spatial uniformity of natural targets will impose a practical limit on the achievable accuracy. Although this problem can be alleviated—for example, by sampling channels simultaneously (as in a grating spectrograph with a detector array) rather than sequentially (as in a filter-wheel radiometer)—it remains a significant source of error.
High radiometric precision also requires exceptional mechanical and thermal stability of the instrumentation because the transmittance of the atmosphere changes rapidly in the neighborhood of absorption lines, so any frequency drift appears as noise.
Doppler frequency shifts caused by the relative motion of planet and satellite become important, requiring precise knowledge of both the satellite motion and the scan direction of the instrument.
Despite the apparent difficulties, there is reason for optimism. A spaceborne instrument to measure pressure would not be a stand-alone instrument but rather would share a common platform with instruments to detect both cloud and aerosol. The radiometric precision and stability requirements are tight but are not unattainable, as has been demonstrated in laboratory experiments by O’Brien et al. (1997). Last, advances in global positioning system (GPS) technology allow the position, velocity, and orientation of space platforms to be determined with sufficient precision.
In an earlier paper, O’Brien et al. (1997) demonstrated experimentally that surface pressure could be tracked with accuracy of approximately 0.1 kPa from O2 A-band spectra acquired in a controlled laboratory experiment. This paper reports airborne experiments with the same spectrograph conducted over the ocean to the south of Adelaide and over the Timor Sea to the west of Darwin, Australia. The spectrograph was pointed toward sunglint on the ocean surface to maximize the reflected signal, and O2 A-band spectra were acquired as the airplane flew a series of descending and ascending legs. The aims of the experiments were threefold:
to demonstrate that variations of surface reflectance across the footprint on the spectrograph are unlikely to have an adverse effect on retrieval of surface pressure;
to demonstrate that the information needed to resolve reflected and scattered radiance can be found in high-resolution spectra; and
to demonstrate that analysis of O2 A-band spectra allows the pressure altitude of the airplane to be tracked with precision of approximately 0.1 kPa in a wide range of atmospheric conditions, including haze layers below the aircraft and thin cirrus above.
Section 2 outlines the flight profiles, instrumentation, and data reduction procedures. Section 3 outlines a simplified model for the radiance used to interpret the major features of the data. The bulk of the analysis is contained in section 4, in which excellent correlations between predicted and measured pressure altitude show that the objectives of the experiment were met.
2. Instrumentation and data reduction
a. Flight profiles
The experiments were conducted in a Fokker F27 airplane operated for CSIRO by Australian Flight Test Services from bases in Adelaide and Darwin, Australia. Three campaigns were conducted, December 1993, March 1994, and June 1994, with a total of nine flights and 30 hours of data acquisition. The atmospheric conditions varied from crystal clear to sharply defined haze layers and thin cirrus. The sea state varied from glassy calm to choppy with high wind speed. Extensive burning in the dry season in northern Australia provided a bonus for the test flights conducted from Darwin. The aerosol loading was high from the bushfires, and the haze extended hundreds of kilometers out to sea. By descending through sharply defined layers of haze, it was possible to demonstrate clearly the effects of scattering upon the O2 A-band spectrum.
The flight profiles varied from day to day, depending on the weather conditions, but the general strategy was the same for all flights. We climbed to maximum altitude and then made a series of slowly descending legs, followed by level return legs, to a height of 60 m above the surface of the sea. Figures 1 and 2 show the flight plan and profile for 18 December 1993; they are typical of all the flights. Legs AA′, BB′, etc., are the acquisition legs, while A′B, B′C, etc., are the return legs.
The instrument was located in the cabin forward of the propeller on the starboard side of the airplane with its scan direction perpendicular to the fore–aft axis of the airplane. In principle, the instrument could scan from close to nadir (vertically down) to the starboard horizon, but, in practice, the scan angle was fixed at the average solar zenith angle for each flight in order to minimize the number of free variables. The heading for each descending leg was adjusted so that the sun was close to the starboard beam of the airplane, so the spectrograph observed sunglint. In order to vary the relative proportions of reflected and scattered light in the beam analyzed by the spectrograph, the heading of the airplane was adjusted to steer the footprint across the sunglint.
The heading on each return leg was opposite to the descending leg, so the spectrograph was on the opposite side of the airplane to the sun and the signal strength was low because the spectrograph collected radiance backscattered from the sea rather than forward scattered. Data from the return legs were not expected to be useful, but subsequent analysis of the contrast between the descending and return legs allowed the components of the radiance entering the spectrograph to be identified because scattered radiance was more important in the return legs.
Ideally, the experiments should have been conducted at high altitude because the proportion of scattered light in the reflected beam increases with the height of the airplane and because the principal advantage of high spectral resolution is that radiance reflected from the surface is “switched off” in observations deep within the O2 absorption lines. When the mass of O2 below the airplane is low, this advantage is reduced. However, the airplane had a practical ceiling of around 6 km, and vagaries of both the airplane and the instrumentation forced nearly all of the observations to be conducted below 2.5 km. The reason related to the thermal control system for the spectrograph detector that used a flow of cold N2 gas boiled from a dewar of LN2. Although this procedure worked well in the laboratory, it was marginal in the airplane because the flow rate depended on cabin pressure, which was poorly regulated.
b. Instrumentation
The spectrograph is a grating instrument used in the Littrow configuration (Born and Wolf 1975) with a focal ratio of f/2.3 and a focal length of 500 mm. The grating, used in second order, has 900 lines per millimeter and dimensions 220 mm × 200 mm. The objective lens is a Petzval configuration. The detector is a 512-element Reticon photodiode array (RL512S) whose pixels have 100:1 aspect ratio, which matches the entrance slit of the spectrograph. The detector is cooled by nitrogen gas boiled from an LN2 dewar. The instrument is mounted in a vacuum tank, whose temperature is regulated above ambient (typically 38°C) by heater mats that cover approximately 40% of the area of the tank. The temperature of the photodiode array and the tank are regulated with a precision of approximately 0.1°C in the laboratory, but in the airplane this degree of stability was achieved only for short periods. All functions of the instrument are controlled by multitasking software running on a 486 PC. Details of the optical design, mechanical stability, software, and detector electronics can be found in papers by Wilson (1989), Petkovic (1989), Jones (1993), Capizzi (1993), Da Costa (1993), and English (1993).
The orientation of the spectrograph in space was determined from the inertial navigation system (INS) of the airplane. During every integration period, the INS was sampled at its maximum update rate, and means and variances of the INS parameters were computed. Whenever the rms variation of the roll, pitch, or yaw exceeded a threshold (typically 0.2°), the spectral data were rejected. The mean values of these parameters were taken to be representative for the integration.
In addition to the O2 A-band spectrograph, the airplane carried instruments to define the atmospheric state, specifically an EG&G dewpoint sensor, reverse flow and Rosemount temperature sensors, and a ParoScientific Corporation static pressure sensor. A nephelometer was included on the flights from Darwin to monitor the aerosol distribution.
c. Dark current subtraction
Each optical integration with the shutter open was followed immediately by an integration for the same time with the shutter closed to measure the dark charge and fixed pattern noise. The dark integrations were subtracted from the solar integrations. The dark spectra showed long memory of the solar spectrum, a well-known feature of photodiode detectors (Weckler 1967;Vogt et al. 1978). However, modeling of the memory effect showed that the errors were sufficiently small that corrections were not required.
d. Nonlinearity correction
e. Temperature profile
f. Air mass
For every optical integration, the geometrical air mass was computed from the time, latitude, longitude, roll, pitch, surface pressure, and pressure height of the airplane. The time was taken from the PC clock, set to local time before the takeoff; latitude, longitude, roll, and pitch were taken from the INS; the pressure height of the airplane was measured by the ParoScientific Corporation transducer; and the surface pressure was determined by making a standard adjustment to the pressure measured on low-level runs over the sea. The reverse flow temperature sensor was used to map the temperature structure of the atmosphere below the airplane. The pressure weighted mean temperature T̃ was used as a proxy for the mean temperature of the lower atmosphere.
g. Correction for roll and pitch
h. Spectral stability
Mechanical stability of the spectrograph is important because spectral drift caused by mechanical or thermal deformation of the spectrograph is equivalent to amplitude noise. Therefore, the position of the detector relative to the spectrum was monitored throughout all flights by using the major absorption lines in the solar spectrum as the reference. The principle underlying the method is that a frequency shift δν of the A-band spectrum produces a phase shift in its Fourier transform that is linear in δν. Thus, by monitoring the phase of the Fourier transform (relative to the Fourier transform of a reference spectrum), the amplitude of any frequency shifts can be deduced.
Tests were conducted to check whether the absolute spectral calibration of the spectrograph was stable between flights. Generally, the procedure involved comparing spectra from different flights in the neighborhood of features where the spectrum changed rapidly.
The first check simply looked for gross changes in the spectrum, such as those that occasionally occurred as a result of strong thermal cycling of the detector when the cooling system was disconnected to change the LN2 dewar. A second test observed a time series of quartz halogen calibration lamp spectra, C(1), C(2), . . . , C(m). The spectra were normalized and corrected for varying illumination and pixel sensitivity. The resulting scatter in pixels provided a useful measure of other highly undesirable effects, such as contamination of the surface of the cold detector. Subsequent analysis showed that the principal causes of variation were thermal displacement of the detector and changing illumination caused by movement of the calibration lamp. Although the displacements were very small, the accuracy requirement of the experiments was so high that all corrupted data were rejected.
3. Radiance model
The radiance Iν at the airplane is a complex function of surface reflectance, solar and view zenith angles, airplane altitude, profiles of pressure, temperature and scatterers, and atmospheric composition. For an instrument in space, only the geometrical factors would be well known. Surface reflectance varies abruptly, depending on wind gusts and surface slicks. Profiles of scatterers and their optical properties generally will be known only poorly, and the temperature profile will rarely be known with accuracy better than 1 K, particularly within the planetary boundary layer. Although data for these variables might be obtainable from other sources, they are unlikely to have sufficient accuracy unless they are collocated and simultaneous. Therefore, it is likely that surface pressure would have to be deduced from data consisting of O2 A-band spectra, an estimate of the mean temperature of the lower atmosphere, and a near-simultaneous estimate of precipitable water.
The role of water vapor requires additional comment because the far wings of water vapor absorption lines in the 11 032 cm−1 band, and in the higher-frequency bands beginning at 13 653 cm−1, extend into the O2 A band. However, the wings are without structure in the O2 A band and so do not affect the ratio of transmittances at frequencies within the O2 A band. Furthermore, the level of water vapor absorption is sufficiently low that it can be estimated with adequate accuracy from current satellite observations of precipitable water derived from passive microwave radiometers. A similar conclusion applies to the effect of water vapor on the mixing ratio of O2. For example, in an atmospheric column with total mass of approximately 104 kg m−2 and precipitable water of approximately 25 kg m−2, the precipitable water need only be known with accuracy ±40% in order to achieve 0.1% accuracy in the total column mass.
This section outlines a simple model for radiative transfer in the O2 A band whose purpose is to identify those variables upon which the radiance depends strongly. It emerges that the radiance is in first approximation a function of the adjusted air mass and that the frequency dependence, although weak, does permit differentiation between reflected and scattered radiance. The analysis is based on a simple radiative transfer model developed by O’Brien (1991) for the O2 A band. The model employs only single scattering but allows the relative importance of scattering by molecules and aerosols to be assessed. Details of the model will be reported elsewhere.
Reflected and scattered radiances were computed for ranges of ζ, η, and airplane altitude corresponding to flights in December 1993 and then were convolved with a triangular transfer function comparable in width to that of the spectrograph. The corresponding spectra are plotted in Fig. 3, which shows clearly that the three components of the radiance are independent functions of frequency.
4. Experimental results
a. Tracking the pressure height of the airplane
During acquisition of the data shown in Fig. 5, the sun was near its zenith, so the zenith angle changed slowly, and the brightness of the sea was approximately constant. Consequently, the fraction of scattered radiance in the beam changed only slowly. On completion of the downward spiral, the airplane again climbed to altitude and commenced a second, more rapid descent from 2000 to 60 m above sea level, indicated in Fig. 1 by leg EE′. On this descent the sea was darker, so the fraction of scattered radiance was higher. Figure 8 compares the results from the two descents. Both descents lead to a linear relation between
These results show that a simple instrument staring at the glint with a single broadband channel can track surface pressure and pressure height with precision close to 0.1 kPa, but the accuracy is unlikely to be better than 2.0 kPa. The errors incurred with such an instrument would depend on the degree of scattering in the atmosphere. We conclude that a two-channel instrument would not provide the pressure data required by the meteorological community. This confirms the theoretical studies by Mitchell (1987) and Mitchell and O’Brien (1987).
b. Sensitivity of the spectrum to scattering
The question arises whether high-resolution spectra of the O2 A band contain additional information that will allow reflected and scattered radiance components to be resolved. In this section we will compare spectra on the acquistion legs AA′, BB′, etc., when the spectrograph observed sunglint, and the return legs A′B, B′C, etc., when the spectrograph observed backscattered radiance, to show that narrowband and broadband channels are not simply correlated and hence contain different information. The important distinction between the acquisition and return legs is that scattered radiance is a larger fraction of the total on the return legs because less radiance is reflected from the sea. The effect is small but real and is consistent with predictions of radiative transfer models. In a later section we will quantify the effect by showing that a combination of low- and high-resolution spectral measurements can account for the variance in radiance under a wide range of atmospheric conditions.
Figure 10 shows channel ratios
On the lowest-level return leg, C′D, represented by the cluster of points on the left of the plot, ratios
X X X X On return leg A′B, corresponding to the clusters on the right, the airplane was higher and the air mass was larger. These clusters lie above the line for
X X On return leg B′C near the middle of the plot, the clusters lie almost on top of the acquisition leg lines.
Quite different behavior is apparent in the ratio
The resolution of this conflict lies in diffuse illumination of the sea surface. When the airplane is low, the only scattering effects of importance are those that occur in the downward flux illuminating the sea. Channels
When
Figure 14 shows the predictions of the single scattering model for ratio
Figure 14 shows that the predictions of the single scattering model are consistent with the data in Fig. 12. As the airplane descends to just above the sea surface, the contribution of scattered light in the reflected beam must reduce to zero, so the data on the return legs converge toward the curve obtained on acquisition legs where reflected radiance dominates.
It is important to note the dominance of the reflected radiance in these simulations. This is principally a consequence of the high solar zenith angle and the low altitude of the airplane. Only on the return legs is the surface reflection coefficient so small that scattered radiance becomes significant. An instrument deployed in space would have to operate with higher air mass, and the scattered radiance would be much larger.
c. Surface reflectance test
The airplane was taken down to an altitude of 60 m above sea level, and spectra were acquired over a 4-min period. The airplane was less stable at this altitude, partly because the autopilot had been disabled and partly because turbulence was higher near the surface. However, variations in roll and pitch were less important because the ray path below the plane was so short. Therefore, variations in channel ratios measured on these legs were caused by instrumental noise and surface reflectance changes. Temperature variations were secondary over such a short period.
To assess the impact of fluctuations in surface reflectance on the accuracy of pressure retrievals, surface pressure was deduced from the broadband channel ratio
d. Pressure altitude predictions
The retrieval algorithm ultimately must be empirical because the uncertainties in the spectrograph transfer function and in the O2 absorption line parameters are too large to permit accurate modeling of the spectrum. The role of physically based models and radiative transfer codes is to understand the impact on the spectrum of factors such as aerosols and temperature profile variations. In this section we examine retrievals based on two models: model A and model B. The first model uses only the broadband channels
In this section we will apply both models to data from the Darwin flights. Although the models are simple, the results are impressive. Model B consistently corrects for changing surface reflectance and scattering in haze layers, whereas model A consistently fails.
Figures 15–18 show for each flight the predicted (p̂) and observed (p) values of the airplane pressure altitude. The data were selected using the criteria described in section 2. The plots are presented in pairs—the first shows the prediction of the simple model A, and the second shows the prediction of model B, which uses both channels
Model B is clearly superior in all cases; the rms error in the predicted pressure altitude is shown in Table 1. The target error of 0.1 kPa was met on three of the four days, despite instrument stability problems on 9 June 1994. Also shown in Table 1 are the ranges of air masses for the flights. Higher air masses were encountered in the Darwin flights than in the Adelaide flights primarily because the sun was lower. On 9 and 12 June 1994, the solar zenith angle at times exceeded 60°, yet the rms pressure errors remained small. This is important because the coverage of a satellite-borne instrument increases significantly if the solar zenith can be extended beyond 60° and because the Fresnel reflection coefficient increases toward the horizon, thereby increasing the signal strength (O’Brien and Mitchell 1990). Therefore, the successful application of the analysis to data where the solar zenith exceeded 60° is very encouraging.
1) Effect of overlying cirrus
While there is a wealth of information to be gained from a close study of all the flights, we will focus on just two legs in order to highlight the most important features.
Figure 17 shows the prediction of model A for the airplane pressure altitude on legs BB′ and CC′ on 11 June 1994. At pressure altitude p ≈ 90.0 kPa, there is a large jump in predicted pressure altitude p̂ between leg BB′ and leg CC′. The adjusted air mass has a corresponding jump, caused only by changes in the positions of the sun and the airplane, because the airplane flew at constant height on the return legs. The solar zenith angle was higher at the start of leg CC′, so the surface reflectance should have increased, given constant surface roughness. However, the integration time does not show any significant change between the end of leg BB′ and the start of leg CC′, so the radiance at the entrance to the spectrograph did not change significantly between legs. As the surface-reflected radiance is the dominant component, there are only two possibilities: either surface roughness increased, reducing the brightness of the sea to its value with the lower solar zenith angle, or the intensity of the solar beam was reduced by thin cirrus or thick smoke haze. Simulations using the single scattering model and the results of the first flight trials in Adelaide show that the channel ratio
Less obvious in Fig. 17 is the change in slope at pressure altitude p ≈ 93.5 kPa. This feature can be explained more easily because the nephelometer and temperature traces show that a haze layer was sitting on top of a weak inversion at 93.5 kPa. When the airplane was above the haze layer, radiance was scattered directly to the instrument, but once below the layer, the instrument detected only a small increase in diffuse radiance reflected from the sea.
Both the step and the kink are eliminated in the model B fit to the data, as shown in Fig. 17. Only O2 A-band spectra have been used; nephelometer data were used to corroborate the presence and optical thickness of the haze layers but were not used in the retrieval procedure. All spectrograph channels respond to the same set of stimuli, namely, the scattering properties of the atmosphere and the reflectance of the surface. That the steps and kinks can be eliminated shows that information about haze, cirrus, and surface brightness has been encoded into the A-band channels. The retrieval procedure attempts to decode this information.
2) Descent through a haze layer
On 12 June 1994, the airplane descended through a sharply defined haze layer trapped above a 1.5°C temperature inversion. The profile of aerosol scattering coefficient is shown in Fig. 20. Above the layer the spectrograph received photons reflected from the sea and from the haze layer. However, once below the layer, the effect of the layer was to increase very slightly the diffuse radiance on the sea surface. The A-band spectrum responded to these regimes in exactly the way expected from the analysis. When model A is applied to the A-band data, the correlation between measured and retrieved pressure altitude is linear below and above the haze layer, but the two lines join with a kink at the haze layer (Fig. 18). Analysis using model B eliminates the kink, so that the measured and retrieved pressure altitudes are closely correlated (Fig. 18) both above and below the layer.
Note that the data analysis was not told that a layer was present at p = 88.0 kPa. The nephelometer data were used only to verify the presence of the haze layer.
3) Comparison of data from different days
The absolute calibration of the spectrograph was not stable from flight to flight, primarily because the detector settled into a new position when it was cycled through a wide temperature range during exchange of the LN2 dewar. Consequently, it is difficult to test the accuracy of calibrations determined on one day with data acquired on another. However, between the flights on 10 and 11 June 1994, the spectrograph appeared to be stable, gauged by the relative position of features of the O2 A-band spectrum on the detector. Data from these days were used in three variations of the calibration procedure:
a single calibration using the data from both days,
predictions on 11 June using a calibration on 10 June, and
predictions on 10 June using a calibration on 11 June.
At present it is not known how wide a range of atmospheric conditions can be encompassed with a single set of calibration coefficients. It is anticipated that a satellite instrument would require auxiliary data to help determine the approximate atmospheric state and that calibration coefficients appropriate to the state would be selected. In the present experiments, the auxiliary data from the nephelometer were used only for corroborating the presence of aerosol layers, not to help define the calibration coefficients.
5. Conclusions
In this paper, high spectral resolution measurements in the oxygen A band have been analyzed with a view to the retrieval of surface presure. Although a simple, broadband, two-channel instrument staring continuously at the sunglint is attractive for a number of reasons, we have shown that the absolute accuracy of such an instrument is unlikely to be better than 2.0 kPa and therefore would not meet the needs of the meteorological community. By contrast, the analysis confirms that the high-resolution spectra do contain information sensitive to the fraction of scattered radiance and that this additional information can be used to enhance the accuracy of pressure retrievals. This is in accordance with the predictions made by Mitchell and O’Brien (1987).
Using a simple linear model for the adjusted air mass in terms of channels with steadily increasing absorption, combined with an empirical calibration, we have retrieved the pressure with an accuracy better than 0.1 kPa with the instrument directed toward the sunglint. Retrievals were even obtained with the instrument directed away from the sunglint, where the reflection coefficient is typically less than 0.1. Here the accuracy was approximately 0.5 kPa.
The experiments highlighted the importance of spectral stability for the spectrograph. The prototype employed cryogenic cooling of the detector and was only marginally acceptable because thermal gradients induced when the system cycled between ambient and LN2 temperatures caused mechanical distortions within the instrument. As a result, empirical calibrations determined on one day could be not applied subsequently. However, the vagaries of the instrument could be reduced significantly (if not eliminated) with better design and therefore should not obscure the principal results of the experiment.
Despite the limitations imposed by the instrument, the present analysis has proven the correctness of the initial proposition, that surface pressure can be retrieved from high-resolution A-band spectra with an accuracy of approximately 0.1 kPa. Translation of these results to a space-borne platform is a difficult engineering problem. Nevertheless, many of the problems encountered in the present series of airborne experiments would not be present in the space environment. Further progress will depend in part on a critical reassessment of the impact of satellite-derived surface pressure measurements on the performance of numerical weather prediction models. Other potential applications of A-band data, including cloud-top pressure and aerosol profiling, also need further consideration. In the meantime, analysis of A-band data from the POLDER instrument on ADEOS will provide an opportunity to assess the value of low-resolution A-band data and may help evaluate the potential of higher-resolution measurements in this spectral region.
REFERENCES
Abramowitz, M., and I. A. Stegun, 1965: Handbook of Mathematical Functions. Dover, 1046 pp.
Allam, R. J., and J. T. Houghton, 1988: The direct measurement of geopotential height from orbiting platforms. Meteor. Mag.,117, 13–21.
Born, M., and E. Wolf, 1975: Principles of Optics. 5th ed. Pergamon, 808 pp.
Bréon, F.-M., and S. Bouffiès, 1996: Land surface pressure estimate from measurements in the oxygen A absorption band. J. Appl. Meteor.,35, 69–77.
Capizzi, V., 1993: Mechanical enhancements to the APS configuration. Proc. Eighth National Space Engineering Symp., Brisbane, Australia, Institution of Engineers, 273–280. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
Cox, C., and W. H. Munk, 1954: Statistics of the sea surface derived from sun glitter. J. Mar. Res.,13, 198–227.
Da Costa, G. A., 1993: APS software development. Proc. Eighth National Space Engineering Symp., Brisbane, Australia, The Institution of Engineers, 192–199. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
English, S. A., 1993: APS detector subsystem development. Proc. Eighth National Space Engineering Symp., Brisbane, Australia, The Institution of Engineers, 287–292. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
Jones, D., 1993: New optics for the satellite-borne atmospheric pressure scanner. Proc. Eighth National Space Engineering Symp., Brisbane, Australia, The Institution of Engineers, 281–286. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
Korb, C. L., G. K. Schwemmer, M. Dombrowski, and C. Y. Weng, 1989: Airborne and ground-based lidar measurements of the atmospheric pressure profile. Appl. Opt.,28, 3015–3020.
McClatchey, R. A., R. W. Fenn, J. E. A. Selby, F. E. Volz, and J. S. Garing, 1972: Optical properties of the atmosphere. 3d ed. Environ. Res. Paper 411 AFCRL-72-0497, Air Force Cambridge Research Laboratories, Bedford, MA, 108 pp. [Available from Air Force Cambridge Research Laboratories, L. G. Hanscom Field, Bedford, MA 01730.].
Mitchell, R. M., 1987: The effects of aerosol scattering on remote pressure measurement via oxygen A-band absorption. Int. J. Remote Sens.,8, 1175–1188.
——, and D. M. O’Brien, 1987: Error estimates for passive satellite measurement of surface pressure using absorption in the A band of oxygen. J. Atmos. Sci.,44, 1981–1990.
O’Brien, D. M., 1991: Passive remote sensing of surface pressure. Data Assimilation Systems: Proceedings of the Second Bureau of Meteorology Research Centre Modelling Workshop, J. Jasper, Ed., BMRC Research Report 27, 247–255.
——, and R. M. Mitchell, 1990: Zones of feasibility for retrieval of surface pressure from observations of absorption in the A-band of oxygen. CSIRO Divison of Atmospheric Research, Tech. Rep. 19, 12 pp.
——, S. A. English, and G. A. Da Costa, 1997: High precision, high resolution measurements of absorption in the oxygen A-band. J. Atmos. Oceanic Technol.,14, 105–119.
Peckham, G. E., C. Gatley, and D. A. Flower, 1983: Optimizing a remote sensing instrument to measure atmospheric surface pressure. Int. J. Remote Sens.,4, 465–478.
Petkovic, M., 1989: Atmospheric Pressure Scanner (APS). Proc. Fifth National Space Engineering Symp., Canberra, Australia, The Institution of Engineers, 107–110. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
Vogt, S. S., R. G. Tull, and P. Kelton, 1978: Self-scanned photodiode array: High performance operation in high dispersion astronomical spectrophotometry. Appl. 0pt.,17, 574–592.
Weckler, G. P., 1967: Operation of p–n junction photodetectors in a photon flux integrating mode. IEEE J. Solid-State Circuits,2, 65–73.
Wilson, I. J., 1989: CSIRO Atmospheric Pressure Scanner spectrograph subassembly. Proc. Fifth National Space Engineering Symposium, Canberra, Australia, The Institution of Engineers, 111–115. [Available from The Institution of Engineers, Australia, 11 National Circuit, Barton Act, Australia.].
WMO, cited 1998: Executive council panel of experts on satellites: Final report. [Available online at http://www.wmo.ch/Documents/www/sat/ecsatftp.doc.].
Flight plan on 18 December 1993. The heading for each leg was selected so that the sun would be broadside to the plane. Legs AA′–EE′ are the acquisition legs, while A′B–D′E are the return legs.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Flight profile on 18 December 1993.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Relative amplitudes of the reflected and scattered radiances at the top of the atmosphere for a section of the O2 A band.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Measured (p) and retrieved (p̂) pressure altitude of the airplane for legs AA′–DD′ on 18 December 1993.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Error Δp in retrieved pressure height of the airplane for legs AA′–DD′ on 18 December 1993.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
The horizontal lines define a transmittance bin (tj−1, tj). All pixels that lie between the horizontal lines are assigned to the set Bj. In subsequent analysis of any spectrum, channel
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Channel ratios
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Channel ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Channel ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Correlation between
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Prediction of the channel ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Comparison of predicted (p̂) and observed (p) values of the airplane pressure altitude for flight 1 on 9 June 1994. Model A uses channels
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Comparison of predicted (p̂) and observed (p) values of the airplane pressure altitude for flight 2 on 10 June 1994. Model A uses channels
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Comparison of predicted (p̂) and observed (p) values of the airplane pressure altitude for flight 3 on 11 June 1994. Model A uses channels
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Comparison of predicted (p̂) and observed (p) values of the airplane pressure altitude for flight 4 on 12 June 1994. Model A uses channels
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Channel ratio
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Profile of aerosol scattering coefficient σ (× 106) for leg AB on flight 4 on 12 June 1994.
Citation: Journal of Atmospheric and Oceanic Technology 15, 6; 10.1175/1520-0426(1998)015<1272:AMOAMF>2.0.CO;2
Errors in retrieved airplane pressure altitude for the flights based in Darwin.
Comparison of errors in retrieved pressure altitude of the airplane when training data are taken from different days.
