MACAWS Data Processing by VAD Algorithms
The MACAWS lidar was flown aboard a series of aircraft missions over the western and southwestern United States during May-July of 1996. This document describes the processing and analyzing of line-of-sight (LOS) velocities from the various missions. One objective was to process the data taken during 180 degree aircraft turns into a series of Velocity Azimith Displays (VADs) for the purpose of simulating what one might see from a conically scanning lidar from space. Vertical profiles of the horizontal winds were then computed. Due in part to the high quality of the data, Flight 18 (July 2, 1996) of the 1996 MACAWS missions was selected as a case study to illustrate the algorithm development, data processing, and analysis schemes employed to construct the VADs and compute vertical profiles of winds.
Figures 1 a, b, c show the ground tracks for three individual VADs (1, 3, and 5) taken during Flight 18. VADs 1, 3, and 5 were selected for analysis due to their similar locations. Ancillary information such as aircraft heading, altitude rolls, pitch and drift for each of the three VADs are depicted in Figures 2 a, b, c. Intensities (in dB) of the MACAWS lidar return signal for VADs 1, 3, and 5 are presented in Figures 3 a, b, c using a color-coded z-t (Range gate vs. time) representation. Corresponding LOS velocities are illustrated in Figures 4 a, b, c. It should be noted that, in the case of Figures 3 and 4, range gate 1 is at the altitude of the aircraft while the ground can be seen between range gates 60-70. A coherent cloud or dense aerosol layer can be seen near range gate 40 in parts of VADs 1 and 3. Each VAD also shows fairly strong signal returns and coherent structure between range gate 40 and the ground. Correlated LOS velocities are also evident in Figures 4 for the mean field (range gates 1-15) and between range gates 40 and 60.
The expected or predicted location (range gate) of the ground return for each VAD was determined through simple trigonometry using the number of range gates, range gate width, and elevation angle for each laser shot. This information, along with range and aircraft altitude is shown in Figures 5 a , b, c.
In order to transpose the graphic representation of Figures 3-4, an algorithm was developed which scanned the data beginning at range gate 100 to find the ground using a pre-determined intensity threshold (50 dB in the case of Flight 18). This range gate was then set as the range gate 1 (the ground) and the subsequent original data was referenced to it. This ground finding was deemed successful when the new ground was within 5 range gates of the trigonometrically expected ground range gate. Failure to find a ground return within the 5-range gate limit resulted in the use of the predicted ground range gate for further processing. The "ground" range gates derived from the original data for the entire VADs are displayed in Figures 6 a , b, c. These range gates were subsequently transposed to range gate 1.
In addition to this strategy, the data was also thinned using an iteratively determined intensity threshold (45 dB in the case of Flight 18) which allowed for the optimum filtering of noisy LOS data. The data was also corrected for ground speed effects by subtracting the apparent LOS ground speed from the entire profile in the instances when the newly found ground range gate was within 5 range gates of the predicted ground range gate noted in Figure 5. The resulting corrected LOS velocities for the individual VADs are presented in Figures. 7 a, b, c.
In Figure 7, VAD 1, there is evidence of a broken cloud layer (with a top near range gate 30) over a portion of the aircraft turning area. These clouds appear to be co-located with an aerosol layer capping an inversion. The aerosol layer (in all three VADs) is approximately 4500m thick and is signified by the coherent LOS wind structure between the ground and range gate 30 in most portions of VADs 1, 3, and 5. Between the top of this layer and the flight level, a rather homogeneous, but with a lower backscatter coefficient, layer is detected. While the signal falls off with range from the aircraft, there is evidence of correlated velocity returns throughout that layer. In order to obtain a good wind vector in these types of layers, at least 10-15 range gates may need to be co-processed.
In order to construct a vertical profile of the horizontal wind components for each of the VADs in Figures 7, a multi-variate analysis using a sine wave was fitted to the entire data set (over 300 shots for VADs 1 and 5) of each VAD and for each range gate. The use of the sine fit was validated by employing a sine wave fitting model that could work with incomplete and noisy data sets. This model, which was developed in Visual Basic by SWA (Wood et al., 1993), allows the user to select the best sine fit by iteratively changing the amplitude, phase, and offset of the wave, as well as the intensity threshold filter to be used.
Figures 8 a, b, c provide an example of the Visual Basic sine fit model for range gates 26, 30, and 64 of Flight 18 VAD 1 using the corrected data and an intensity threshold of 45 dB. The Goodness of Fit (GOF) for the three individual range gates (59.2, 39.3, and 67.7, respectively) reflects the somewhat noisy nature of the data. However, as the intensity threshold is increased, a lesser number of "noisy" observations are accepted and the GOF improve significantly. This is illustrated in Figures 9 a, b, c using range gate 64 as an example.
Using 45 dB as the intensity threshold (consistent with earlier filtering), the multi-variate analysis was employed to calculate the horizontal wind components at each range gate for each VAD. The vertical coordinates of the resulting profiles were then converted from range gate to heights above ground. Vertical profiles of the horizontal wind components for Flight 18 VADs 1, 3, and 5 are presented in Figures 10 a , b, c. The coherent structure within each profile, and between the profiles of the individual VADs (particularly between VADs 1 and 5), is obvious and very encouraging. In addition, the agreement between the u and v component wind speeds at the top of the profiles and the wind speed and direction measured aboard the aircraft (Figures 11 a , b, Figures 12 a , b, Figures 13 a , b) lends support to both the MACAWS data and the processing methods that were chosen.
However, one feature from Figures 10 requires an explanation. This is the "missing" data in the middle of the vertical profiles. This feature is present because the curve fitting and the generation of u and v components for each VAD at each range gate was only undertaken when there were at least 10 observations/shots that met the intensity threshold. This was not the case in many instances. On the other hand, this is not to say that there is no meaningful data at these locations. Undertaking this processing and computing the u and v components for a 3-5 range gate wide layer would most likely yield important data. This will be pursued in future research.
List of Figure Captions
Figures 1: Location of flight tracks for VADs 1, 3, and 5 during 1996 MACAWS Mission Flight 18.
Figures 2: Ancillary scanner information for Flight 18 VADs 1, 3, and 5. Time is in decimal format (hundredths of an hour) for plotting purposes.
Figures 3: Range gate vs. time diagram of MACAWS lidar return signal intensity (in dbs) for Flight 18 VADs 1, 3, and 5. Time is in decimal format and range gate 1 is at the aircraft flight altitude.
Figures 4: Range gate vs. time diagram of MACAWS line-of-sight (LOS) wind velocities for Flight 18 VADs 1, 3, and 5. Time is in decimal format and range gate 1 is at the aircraft flight altitude.
Figures 5: Ground range gate location and distance from the aircraft for each lidar shot during Flight 18 VADs 1, 3, and 5 as determined by trigonometric solution. Aircraft altitude and elevation angle is also presented. Time is in decimal format.
Figures 6: Ground range as determined by ground-finding algorithm for Flight 18 VADs 1, 3, and 5. Time is in decimal format.
Figures 7: Corrected LOS velocities as a function of range gate and time for Flight 18 VADs 1, 3, and 5. Time is in decimal format and ground is located at range gate 1.
Figures 8: Visual Basic sine fit model for range gates 26, 30, and 64 of Flight 18 VAD 1. AN intensity threshold of 45 dB was employed.
Figures 9: Visual Basic sine fit model for range gate 64 of Flight 18 VAD 1 determined using intensity thresholds of, respectively, 60.0, 61.0, and 61.6 dB.
Figures 10: Vertical profiles of the horizontal wind components derived from the multi-variate analysis of LOS velocities from each range gate for Flight 18 VADs 1, 3, and 5.
Figures 11: Wind speed and wind direction at aircraft flight level during Flight 18 VAD 1.
Figures 12: Wind speed and wind direction at aircraft flight level during Flight 18 VAD 3.
Figures 13: Wind speed and wind direction at aircraft flight level during Flight 18 VAD 5.
Wood, S.A., G.D. Emmitt,
M. Morris, L. Wood and D. Bai. Space-based Doppler lidar
sampling strategies -- algorithm development and simulated observation experiments. Final
Rept. NASA Contract NAS8-38559, Marshall Space Flight Center, 266 pp., 1993.
Return to SWA Lidar Home Page
Return to SWA Home Page
|This page managed by Sidney A. Wood||Last modified: 28 Jan 1999|