Table 5.2:

NOAA Twin Otter Aircraft Instrumentation Payload for CASES-97:

This was the initial experiment for the newly installed flux system on the Twin Otter. Full details of the instrumentation system can be found in Wellman et al., 1996, however, a few pertinent details will be reviewed here. The turbulence measurement system was basically derived from the Atmospheric Turbulence and Diffusion Division (ATDD) Mobile Flux Platform (MFP) system (Crawford et al., 1990), but included some major changes. Chief among these was the complete dependence on Global Positioning Satellite (GPS) system data to determine not only aircraft position and velocity, but attitude as well. The system does not use an inertial navigation system, gyros, or accelerometers, but relies on 20 Hz position and velocity data and 4-10 Hz attitude data.

Special Instrumentation: Two instruments built in-house warrant special attention. H₂O and CO₂ are measured by an open path, fast response infrared gas analyzer (IRGA) mounted on the nose of the aircraft. This instrument has performed reliably in several field experiments in both terrestrial and marine environments. Noise levels for H₂O and CO₂ concentrations are less than 10 mg/m3 and 300 mg/m3, respectively, for frequencies between .005 and 10 Hz. Full details of this instrument can be found in Auble and Meyers, 1992. Ozone was measured by a dry disk, chemiluminescence system, designed and constructed at ATDD. The reactive disk is coated with a Courmarin dye solution, which fluoresces when combined with ozone. Sensitivity of this instrument degrades with the age of the Coumarin disk, requiring daily in-flight calibration against the slower reference ozone analyzer.

A schematic of fast response instrument location is given in Figure 5. An extension to the standard nose provides space for instrumentation and mounting locations for all fast response instruments. Pressure ports are located on the spherical nose of the extension, in close proximity to the pressure sensors, which are mounted inside the extension. The IRGAs are self-contained and are mounted below the extension. Two net radiometers are mounted on horizontal struts on each side of the nose extension. The fast air temperature sensor is mounted atop the extension in a Rosemount housing. The inlet for the fast ozone analyzers, as well as the LiCor 6262, is located between the two IRGAs, on the bottom of the nose ring. Inlet and pressure tube lengths

Table 1. Fast-response Instruments

were kept as short as possible; pressure tubing is less than 0.5 m and the ozone inlet tube is less than one meter.

GPS position and velocity: Position and velocity are determined using two Novatel 3951 Paccar receivers, one in a ground-based computer serving as a base station, operating at 2 Hz, and the other aboard the aircraft, operating at 20 Hz. Both receivers record the raw pseudorange data from every visible satellite. In post-processing, the aircraft data are differentially corrected, providing position data accurate to about 2 m and velocity data accurate to about 2 cm/s. In quality control tests, velocities have been integrated over four hour flights with errors between the beginning and ending positions less than 20 m.

GPS Attitude: Normally, attitude information (pitch, roll, and heading) is provided at 1 Hz by a Trimble TANS Vector GPS system. Accuracy of pitch, roll, and heading is +/- 0.1 degrees (+/- 2 mrad). During SOS, however, prototype software, provided by the manufacturer, was used to enhance attitude processing, allowing processing of raw data to be shared between the GPS receiver and a dedicated laptop computer. Attitudes were obtained at rates between 4 and 10 Hz. Since the completion of the SOS study, the TANS Vector system has been upgraded to perform at 10 Hz in a stand-alone mode.

Table 2. Slow-response Instruments.

Table 3. Novatel GPS Output (after differential processing).

Table 4. TANS Vector GPS Output.

Data Collection: Data were collected using a commercially available system, the Science

Engineering Associates Model 200 Data Acquisition System, and recorded on 4 mm Digital Audio Tape cartridges. There were four major groups of data collected, the fast-response channels, the slow-response channels, the Novatel GPS data, and the TANS GPS data. The 11 high-speed channels listed in Table 1 were simultaneously sampled at 200 Hz, and block averaged to 40 Hz on post-flight processing. The 28 low-speed channels, most of which are listed in Table 2, were scanned at 2 Hz and block averaged to 1 Hz in realtime. Novatel GPS data were acquired at 20 Hz. These data were acquired in a passthrough mode, being written directly to tape with no processing aboard the aircraft. Data from the TANS Vector GPS instrument were collected laptop computer which, using the prototype software provided by Trimble, shared the computational chores with the CPU in the TANS Vector system. This allowed the aircraft attitude to be determined at frequencies ranging from 4 to 10 Hz, depending on conditions and the particular computer configuration employed. Most of the attitude data during SOS were acquired at 4 Hz. In addition to airborne data collection, a GPS reference station was operated at the Nashville airport. This system used a Novatel GPS and collected raw pseudoranges from every visible satellite at 2 Hz.

Data processing consisted of several steps. First the data set was transferred to hard disks on the PC based network. After the four types of data outlined above were separated into individual files, the Novatel GPS pseudoranges were differentially corrected by the ground station pseudorange errors and accurate position and velocity solutions calculated. The next step consisted of visual examination of time series data for each instrument, examination of statistics for each instrument, and , where appropriate, analysis of power spectra of the time series assured data quality. Next, the four files were combined, using the appropriate algorithms, to form a 40 Hz file of all data, including a three-dimensional wind vector.

Table 5. Calculated Data (*estimated, depends on site and external influences)

Processing algorithms

Eddy correlation calculations separates each time series into steady state and fluctuating components. Classically this is accomplished in the evaluation of the covariances between the various time series, and is inherently linked to the averaging period over which the covariances are formed. The removed mean (the steady state component) has, by definition, the same averaging period as the covariance averaging period. A more flexible approach is to remove a quasi-steady state component using a high pass filter which may be tuned to the local conditions (i.e. plumes), and to evaluate covariances over the shortest averaging period which produces statistically robust results. Using this method, we can reduce the adverse effect of local plumes on the calculation of fluxes. For fluxes measured near the surface, we high-pass filter the data using a 4 pole Butterworth filter with a 100 second time constant. Covariances are formed from the high pass data, using a 60-second data window. The window is then advanced 20 seconds forward, and the covariances are reevaluated. This process continues through the end of the flight. This technique produces covariances that are not statistically independent from the values calculated immediately before and after them, but provides improved spatial resolution. A coordinate rotation is performed on the covariances so that u is now along the mean wind direction, v is crosswind, and w is zero. We could probably neglect this operation in flat terrain, but it is a requirement for eddy correlation measurements made in complex terrain (McMillen, 1988).

In other cases it is more appropriate to evaluate the covariances along the entire transect (such as when there is little interest in the spatial variation of the fluxes). The fluxes along an entire transect are naturally more robust, and this method was preferred when evaluating flux divergences. Other than the length of the window over which the covariances are evaluated, the techniques are identical. The coordinate rotation is the same.

Quality Control Issues

Raw and processed time series data were visually inspected in both time and frequency domains to ensure overall quality of both the instruments and the data acquisition system. Evidence of spiking of the fast air temperature sensor was noticed, and was found to be caused by interference from radio transmissions by the pilots. Further investigation showed that several other sensors were affected by radio transmissions, but none as severely as air temperature. The spikes were deleted form the time series. Other variables were unaffected by radio interference.

Typical power spectra of the fast channels are shown in Figure 6. The power spectra follow the theoretical ] 2/3 slope in the inertial sub-range varying fidelity. W and u show some damping of the signal at frequencies greater than 5 Hz. T begins to deviate from a �2/3 slope at about 1 Hz, while H₂O shows slight damping at about 10 Hz. The CO₂ spectra shows the onset of noise at about 0.5 Hz. Since this instrument in a a vibration-free environment demonstrates much better signal-to-noise ratio, an improved aircraft mount is currently being developed. The ozone spectra deviates from the �2/3 slope at about 1 Hz. The large spike at 3 Hz was apparently due to resonance in the intake system, since modifications made in the intake since the completion of SOS have removed the spike. As the co-spectra in Figure 14 indicates, the spike is not correlated with the w signal, and occurs at a a frequency where little or no transfer occurs.