ELDORA Data User's Guide for TOGA COARE

Transcription

1 NCAR/TN-408 NCAR TECHNICAL NOTE November 1994 I ELDORA Data User's Guide for TOGA COARE Peter H. Hildebrand Wen-Chau Lee Robert Rilling NCAR/ATD/Remote Sensing Facility Richard Oye NCAR/ATD/Research Data Program ATMOSPHERIC TECHNOLOGY DIVISION NATIONAL CENTER FOR ATMOSPHERIC RESEARCH BOULDER, COLORADO

2 ELDORA Data User's Guide for TOGA COARE Peter H. Hildebrand Wen-Chau Lee Robert Rilling NCAR/ATD/Remote Sensing Facility Richard Oye NCAR/ATD/Research Data Program 23 November, 1994

3 Table of Contents 1. Introd uction ELDORA Operations in TOGA COARE D ata P rocessing Error Sources and Corrections Corrections Applied to Data Half-Nyquist Folding Correction Aircraft Ground Speed Removal and Unfolding Doppler Velocities UNIX File Size Problem for Copies of ELDORA Data Made in Honiara Bias and Uncertainty in INS and Antenna Positioning Converting Data to DORADE and UF Data Compression Analysis Software Scan-plane Data Visualization and Editing Interpolation, Multiple-Doppler Analysis, and Display Obtaining Data, Software, and Answers Appendix A: DORADE Format Overview Appendix B: Radar Coordinate Transformation i

4 List of Tables Table 1. ELDORA Radar Characteristics as Operated in TOGA COARE...2 Table 2. Summary of ELDORA Operations in TOGA COARE... 4 Table 3. Definitions of Five Classes of TOGA COARE Convections... 5 Table 4. ELDORA Parameters Provided for TOGA COARE Table 5. ELDORA Correction Factor Evaluation. The Correction Factors Are Evaluated for Two or More Groups of Scans on Five Flights in February Each Group of Scans Contains 25 or More Individual Scans Table 6. List of Persons to Contact Regarding ELDORA Data List of Figures Figure 1. Schematic Diagram Showing How the Half-Nyquist Folding Problem Occurred During TOGA COARE due to Improper Averaging of Velocities from Two Frequencies as Described in the Text... 8 ii

5 1. Introduction The purpose of this technical note is to provide a concise User's Guide for the National Center for Atmospheric Research (NCAR) ELectra DOppler RAdar (ELDORA) data collected during the Tropical Ocean and Global Atmosphere Program Coupled Ocean-Atmosphere Response Experiment (TOGA COARE). This guide includes 1) a description of the radar system configuration, 2) characteristics of the sampling strategy, and 3) the ELDORA operations in TOGA COARE. The major portion of this User's Guide discusses the corrections applied to the data, the ELDORA data formats available, the data processing steps used, analysis routines available, and the means of obtaining data. Instructions for obtaining additional information are provided, including how to obtain assistance and how to access on-line documents concerning updates to data and/or calibrations. The appendices present a brief description of the data format and a summary of the coordinate transformation matrix. The ELDORA radar as configured for TOGA COARE and mounted on NCAR's Electra aircraft can sample storm hydrometeor motions and reflectivities over a domain extending km from the flight track. The ELDORA radar (Hildebrand et al. 1994) is an X-band Doppler radar with 35 kw peak power, a pulsed dual-frequency waveform, a beamwidth of 1.8, = 40 db antenna gain, and a horizontal polarization when the antennas are pointed horizontally (see specifications in Table 1). The radar employs dual flat-plate antennas with fore- and aft-pointing radar beams. These beams are scanned in cones which are tilted by ± fore and aft of a plane which is perpendicular to the aircraft heading. This technique was developed to enable the radar to collect dual-doppler information as the aircraft passes by or through atmospheric storms. The radar observations are then used to reconstruct storm structure and kinematics. Data from the ELDORA operations in TOGA COARE have been carefully reviewed and are now available for research use. Although some signal processor problems limited operations to two transmit frequencies during TOGA COARE, the data still meet the ELDORA design goals for the absence of statistical sampling noise. The data generally have smooth, noise-free reflectivity and velocity fields which are much more like ground based radar data than previously obtained airborne radar data. Achieving this level of data quality was a major design goal for the ELDORA development and the primary motivation for using a multiple frequency transmitted waveform. 2. ELDORA Operations in TOGA COARE During January and February 1993 the ELDORA airborne Doppler radar was operated on the NCAR Electra aircraft in support of TOGA COARE (Table 2). The radar was installed on the aircraft in December 1992 during the break between the second and third Intensive Operational Periods (IOPs). Thereafter, ELDORA operated on 13 out of 1

6 Characteristic TOGA COARE Design Goal Wavelength (cm) Transmit Frequency (GHz) 9.35, Beamwidth (deg, circular) Antenna Gain (db: fore/aft antennas) 38.75/ Polarization (antenna horizontal horizontal horizontal Sidelobes (d) -23 to Beam tilt angle (fore or aft, deg) Antenna spin axis parallel to: heading heading Antenna rotation rate (deg/s) Dwell time (ms) Rotational sampling interval (deg) 1 1 Peak Transmit Power (kw) Minimum Detectable Signal at 10 km (dbz) Receiver Bandwidth (MHz) 2 2 Receiver Temperature at Antenna ( K) <600 <600 Pulse Repetition Frequency (pps) Unambiguous Range (km) Unambiguous Velocity (m/s) Number of Radars, 1 (switched) 2 (fore/aft) Number of Xmit Frequencies per Radar 2 3. Pulse Chip Length (pis) Range Averaging (chips) 1-4 Total Gate Length (km) Elevation Step (deg) 1 1 Along-Track Beam Spacing (km) Table 1. ELDORA Radar Characteristics as Operated in TOGA COARE. the 16 NCAR Electra flights during IOP's #3 and #4. The first 7 flights served as ELDORA check-out missions during TOGA COARE boundary layer studies. The remaining 6 flights provided ELDORA observations of TOGA COARE Class 1-3 convection. The TOGA COARE convection was categorized according to the area coverage of the 208K IR cloud top temperature (see Table 3). Four of these 6 flights included coordinated data collection with the National Oceanic and Atmospheric Administration (NOAA) P-3s and other TOGA COARE aircraft. The start of ELDORA support of TOGA COARE was delayed due to severe reliability problems with the transmitter high power amplifiers (HPAs). Partial repairs of the HPA units were completed in time for the start of IOP #3 and produced the needed high quality radar transmissions, but only with the reliability to operate for about 20 hours between major HPA failures. Since just two HPA units were available and the TOGA 2

7 COARE research was expected to extend considerably longer than 20 hours, the radar was operated with only a single transmitter and a waveguide switch. In this configuration, the single transmitter switched between the fore and aft antennas on a scan-by-scan basis. (The dual-transmitter ELDORA design is intended to provide simultaneous 360 observations from fore and aft antennas.) In the early flights (missions 23-31) the switching was primarily performed on a complete 360 scan basis. In the latter flights (missions 33-38) the radar was principally switched every 180 to provide observations on one side of the aircraft track only but with higher resolution. In TOGA COARE, ELDORA operated with 650 m along-track resolution for the 180 scans and 1300 m along-track resolution for the full 360 scans. The best ELDORA radar data were collected on TOGA COARE missions 33-38, the last six Electra flights. These missions included Class 1-3 convection, and included flights with the NOAA P-3s and other TOGA COARE aircraft. The earlier Electra flights (missions 23-31) occurred during boundary layer missions with Class 0-1 convection and provided the opportunity for ELDORA's initial test flights, producing only a small quantity of ELDORA data. During this period three problems in the data system and the operational procedures were solved. First, a calculation problem which produced incorrect velocity values at Nyquist folds was discovered between the first and second ELDORA flights (flights 23 and 24). (See Section 3.1 on page 5 for a complete description of the problem and its solution.) Second, a disabling intermittence developed in the high bandwidth data cable between the digital signal processor and the rest of the data system (displays, recording, etc.). This problem was finally located and repaired after mission 31. The data system then ran reliably for the rest of TOGA COARE. Finally, there were the normal variety of operational quirks which any new system has on its first outing. These problems were principally corrected during missions Data Processing Each beam of ELDORA data has the following attributes: aircraft location (latitude, longitude, altitude), aircraft velocity (with respect to both the air and the ground), time, and azimuth and elevation angles (see Lee et al. 1994a for definitions and descriptions of the airborne Doppler radar geometry and terminology).. Similar to groundbased systems, azimuth is the horizontal pointing angle of the radar beam measured clockwise looking down from north, and elevation is the vertical angle above the horizontal plane on which azimuth is measured. The rotation angle of the antenna about the spin axis (positive in a clockwise sense looking from the tail to the nose of the aircraft) and the tilt angle of the scan cones fore and aft of a plane normal to the aircraft heading are recorded, as is the aircraft attitude information needed for calculating all angles. In the delivered data set, the small errors inherent in these measurements have been corrected by using radar measurements of ground reflectivity and velocity (see Section 3.1.5). 3

8 TC Date Mission Electra ELDORA Other Acft Ops Flt# GMT Class' Flt# Times 2 Date h Times h N42RF N43RF DC-8 ER-2 C130 C Nov-Dec Conv Jan /1 BL x x BL/Conv /13 Jan BL /14 Jan x BL /15 Jan x I1BL /17 Jan x x BL /18 Jan x x x BL /19 Jan x x x x OBL /27 Jan x OBL /28 Jan x OBL /29 Jan x OBL IFeb Conv /5 Feb Conv Feb x Conv Feb /3 Conv /11 Feb x lconv /18Feb x Conv /19 Feb O BL 19 Feb Conv 20 Feb x x Conv 22 Feb x x 1 From TOGA COARE Turboprop Mission Summary At-A-Glance (Marks and Smull 1993, personal communication) 2 From NCAR/RAF TOGA COARE Project Documentation Summary Table 2. Summary of ELDORA Operations in TOGA COARE. 4

9 CLASS AREA OF TEMPERATURE < 208K 0 INone -I <6000km 2 II km 2 III km 2 IV >60000 km 2 Table 3. Definitions of Five Classes of TOGA COARE Convections. The ELDORA data are written in ELDORA FIELD format (Walther and Lee 1994), which is an extension of the DOppler RAdar Data Exchange (DORADE) format (see Appendix A and Lee et al. 1994b). DORADE format (version 1) is the preferred exchange format for TOGA COARE ELDORA data. The DORADE format supports data compression, reducing the data amount by 60%, compared with the data amount on the Universal Format (UF) (Barnes 1980). Note that there is some loss of information (particularly concerning correction terms) in the conversion to UF (see section 4.1). It is, therefore, hoped that DORADE is the exchange format of choice for the future. Users are advised to start considering ways to modify their software to make use of DORADE. Since new software to handle DORADE format is still under development, TOGA COARE ELDORA data are available in both DORADE and UF. DORADE and UF format data have been staged to the NCAR Mass Store System (MSS), and are available for remote accessing of the data. These data are also available directly from the Atmospheric Technology Division (ATD). Users will still be able to use the existing post-analysis software (i.e. Perusal, Editor, REORDER, and CEDRIC) to analyze ELDORA data. To date, both REORDER and SPRINT software have been enhanced to ingest DORADE data. A UNIX X-Window based radar perusal and editor package (SOLO) is currently being developed at NCAR. This software will make extensive use of DORADE. Fortran programs with C I/O modules are available from the Remote Sensing Facility (RSF) at NCAR/ATD which can demonstrate methods of access to DORADE format data. These programs can be requested, used and modified freely by the users, but are NOT intended to be an ATD-supported software. The balance of this section presents the data processing steps which were used to convert data from ELDORA FIELD format into DORADE and UF. In addition, the procedures to obtain ELDORA data from the NCAR MSS or from NCAR/ATD are discussed. 3.1 Error Sources and Corrections The detailed descriptions of the known error sources will be discussed in this subsection. The solutions and the corresponding corrections are also provided. 5

10 DORADE NAME UF NAME Parameter Description DBZ DB Reflectivity factor NCP NC Normalized Coherent Power SW SW Spectral Width VR VR Raw Velocity as recorded by the ELDORA data system VN VN Raw Velocity with a correction for the error at velocity folds introduced by averaging multi-frequency velocities VG VG Velocity corrected for aircraft ground speed and other aircraft motions VU VU Unfolded Velocity, derived during objective batch processing from VG Table 4. ELDORA Parameters Provided for TOGA COARE Corrections Applied to Data The following errors in ELDORA FIELD tape housekeeping have been identified. These new housekeeping values have been used in producing the final data set (DORADE or UF), and should be transparent to the users. However, any users who access data from ELDORA field tapes (or who obtained UF data from RSF in Honiara) should be aware of these problems. 1. The transmitted frequencies in the field tape header were incorrect. The actual transmitted frequencies were 9.35 and 9.45 GHz. The average frequency was 9.40 GHz. The effect is that the Nyquist velocity in the header and the Doppler velocity values on the tape were off approximately 1%. The corrected scale factor for the velocity is The scale factors for normalized coherent power (NCP) and spectral width (SW) in the header were incorrect in the field tape. The corrected values are for NCP and 7.95 for SW. NOTE that the UF and DORADE data contain incorrect SW scale factor (4.42 instead of 7.95). The SW value is about 1.8 times larger than its actual value. 3. The reflectivities for the fore and aft radar beams were computed using only one radar equation and radar constant during TOGA COARE. The reflectivity fields have been recalibrated to account for different antenna gains and different transmitted power from the two HPAs used. The corrected antenna gains are db for the fore antenna and db for the aft antenna. 4. The Julian day recorded in the field tape ray header was sometimes incorrect. The date in the volume header was correct; however, month = 0 and month = 1 both indicated January. 6

11 5. The horizontal winds (u and v) in the platform info block were incorrect and have been corrected. 6. The antenna H-plane angle used in the calculations of rotation angle was incorrect. This angle affects rotation angle, azimuth, and elevation. The rotation angle for the aft antenna was corrected by The correct tilt angle for the fore and aft antennas was + and , respectively. 7. The roll angle recorded in the platform info block was actually the roll angle rate. These values are generally within 1 of the Research Aviation Facility (RAF) Aircraft Data System (ADS) data; however, the difference can become large when the aircraft turns. The correct roll angles from the ADS data tape have been merged into the final ELDORA data Half-Nyquist Folding Correction During TOGA COARE ELDORA operated at the two transmit frequencies of 9.35 and 9.45 GHz, for which the Nyquist velocities are at and m/s, respectively. During TOGA COARE the Doppler velocity was calculated by averaging pulse-pair mean velocity estimates for each separate frequency, rather than by averaging the real and imaginary components of the vector and then taking the arctangent to obtain the Doppler velocity estimate. Due to the different Nyquist velocities of and m/s, erroneous near-zero velocities occurred after the Doppler velocity folds at 9.35 GHz Nyquist velocity, but not folds at 9.45 GHz Nyquist velocity. A schematic diagram showing how the problem occurred near the Nyquist velocity is illustrated above in Figure 1. These near-zero velocities always occurred between positive and negative Nyquist velocity. Using the above characteristic, the data have been corrected by identifying the problematic values and then adding or subtracting half a Nyquist interval (depending on the sign of the original Doppler velocities, VR). A new velocity field, VN, has been created after unfolding these half- Nyquist-folded velocities. In creating VN, VR was thresholded by normalized coherent power of 0.33 to remove background noise -- a critical step for the success of this scheme. Very few outliers escape this scheme undetected. These few erroneous values remain in the data and must be deleted interactively using a data editor Aircraft Ground Speed Removal and Unfolding Doppler Velocities The ELDORA antennas scan about an axis parallel to the aircraft heading, at tilt angles fore and aft of the plane perpendicular to the fuselage. As a result, the observed Doppler velocities contain the velocity components not only from hydrometeors but also from the aircraft's flight speed. The equation to compute the component of the aircraft's flight speed in the 7

12 Averaging Problems Near Nyquist Velocities E 5o0 > 0 Q0I Gate Number Figure 1. Schematic Diagram Showing How the Half-Nyquist Folding Problem Occurred During TOGA COARE due to Improper Averaging of Velocities from Two Frequencies as Described in the Text. beam-viewing direction has been documented in Lee et al. (1994a). At a ±18.5 tilt angle and a120 m/s aircraft ground speed, the component of the aircraft ground speed along the radar beam is about ±38 m/s. Multiple folding in Doppler velocities thus occurs, considering that the Nyquist velocity for the ELDORA system was =16 m/s during TOGA COARE. A new velocity field, VG has been created by subtracting the aircraft-ground speed component from VN. Finally, the VG velocity values were unfolded using the Bargen and Brown (1980) "B"- algorithm. This process unfolds each gate separately, using the along-beam component of the insitu wind as a reference to unfold the first gate. The following gates are then unfolded, using the unfolded Doppler velocity from a running average of (up to) the 7 previous gates as a template. The resultant unfolded Doppler velocity has a field name VU. Normally, the unfolding and aircraft ground speed removal are both accomplished in the same step. All velocity fields (VR, VN, VG, and VU) are provided to the users of ELDORA data (Table 4). 8

13 3.1.4 UNIX File Size Problem for Copies of ELDORA Data Made in Honiara The tape copying routine that the RSF distributed in Honiara had a default file size limit of 2 Gbytes. This problem caused the tape copying to stop at the 2 Gbyte mark and exit prematurely. All ELDORA field format data past this point which were copied in Honiara were therefore not copied, and are missing from "copy" tapes. Due to this and the above-mentioned corrections made to the ELDORA FIELD tape data, RSF strongly recommends that users obtain a DORADE or UF tape which contains the most recent updates Bias and Uncertainty in INS and Antenna Positioning The accuracy of the unfolded Doppler velocities, and their mapping into an earth-relative coordinate system, is critically influenced by biases in the Inertial Navigation System (INS) and by other antenna positioning errors such as mounting errors, aircraft altitude error, and radar range delay error. The major INS error is the 90 min INS Schuler oscillation (Rodi et al. 1991) which produces an error in the aircraft location (longitude and latitude) as well as ground speed. This bias can be corrected by using the reliable but relatively infrequent GPS data as a template. The longitude and latitude information in both the DORADE and UF format have been adjusted accordingly. The stability of the raw recorded aircraft attitudes are within 'the original design expectations. The drift angle has the largest uncertainty among all INS variables. If the INS variables and the antenna positions were measured perfectly, and the aircraft ground speed and unfolding are performed correctly, the ocean surface should appear to be not moving (it is assumed that the ocean current is negligible in normal conditions). The intersection of the ocean surface and a conical helix of radar beams is nearly a hyperbola. The ocean surface would appear level if the hyperbola is projected to a plane normal to the fuselage properly. Given the expected uncertainties in the raw aircraft attitude and positioning data it is common to find that even after performing all the data processing steps, the sea surface is not quite flat and level and the velocity of the sea surface is not quite zero. Using the range and residual velocity of the ocean surface as constraints, a variational technique has been derived (Testud et al. 1994) to correct systematic biases in drift, pitch, rotation angle, aircraft ground speed, aircraft pressure altitude, range delay, and tilt angle. Errors in these values can be deduced by evaluating: 1) the predicted versus measured range error of the surface range gate for each beam in the sweep, and 2) the bias of the surface velocities away from zero. The biases produced by different errors result in different signatures in the velocity and the range residuals within a sweep. The analytical technique of Testud et al. (1994) can thereby separate the biases contributed by the different parameters. An iterative process combines the unfolding, aircraft ground speed removal, and the INS bias removal to calculate a set of correction factors which -- in least squares sense-- minimizes the velocity and range residuals along a sweep. 9

14 The correction factors were evaluated using from 14 periods encompassing the ELDORA flights during TOGA COARE IOP #4. Error Sources and Corrections summarizes the findings, giving the average correction factor for a number of important variables: range to the first gate (fore and aft radar), pitch, drift, rotation angle, aircraft altitude and aircraft horizontal velocity. The correction factors in the range to the center of the first gate (range delay, RO) for the fore and aft radars were nearly constant for all time periods reviewed. The variability in the range delay error is considerably less than a gate length, which is the upper limit of the uncertainty in range residual. These values agree with values estimated from measurements in the field. The pitch and rotation angle have small, stable biases which are within expectations. The most unstable parameter is the drift bias, which, due to the measurement error in the heading, can vary within about ±+1. The measured drift bias variability of ±-0.7 is consistent with this expected measurement error. This drift bias variability should be expected to introduce variable errors in the aircraft ground speed of =1 m/s. The flight-to-flight deviations among all other parameters are negligible except for the aircraft pressure altitude, where day-to-day variations are expected due to the changes in surface pressure. Our analysis indicates that a single set of corrections for the entire TOGA COARE ELDORA data set should account for most errors in navigation of the data. These corrections have been derived and applied to the aircraft's pitch, roll, and radar range delay. Small additional corrections should be added by users in analysis to correct for drift and subsequent aircraft ground velocity error. Users desiring to correct for these residual errors should expect to fine-tune their data set with help from the FORTRAN code provided by RSF for analyzing the above biases. The final data are generated using the average correction factors obtained in the iterative process. For TOGA COARE data the correction factors are: * 2.3 for rotation angle * for pitch * -10 m for the aircraft altitude * 10 m for the range delay 3.2 Converting Data to DORADE and UF The conversion of data from ELDORA FIELD Format to DORADE or UF involves corrections to both the tape-header and the data. In addition to the standard ELDORA variables such as reflectivity factor (DBZ or DB), normalized coherent power (NCP or NC), spectral width (SW), and raw velocity (VR), three new velocity fields are presented on the tape. These parameters are defined in Table 4. By providing all the fields, users can easily reformulate any of the corrections which have been applied to the data, should they wish to make their own corrections, or if the correction factors are updated by RSF. 10

15 Date Time #Scans RO fore RO aft Pitch Drift Rotation Altitude Horizontal Angle Velocity (GMT) (m) (m) (deg) (deg) (deg) (m) (m/s) 06 Feb 93 18:47-18: Feb 93' 19:48-19: Feb 93 16:08-16: Feb 93 18:34-18: Feb :40-19: Feb 93 18:46-19: Feb :37-20: Feb 93 20:53-20: Feb :06-23: Feb 93 20:08-20: Feb 93 20:30-20: Feb 93 22:19-22: Feb 93 21:28-21: Feb 93 01:00-01: Mean Values Variable corrections between 19:52-19: Aircraft turn and variable corrections around 19: First file of tape: subsequent files have different corrections. 4. Aircraft climbing. Heading changes smoothly from Table 5. ELDORA Correction Factor Evaluation. The Correction Factors Are Evaluated for Two or More Groups of Scans on Five Flights in February Each Group of Scans Contains 17 or More Individual Scans. All the above corrections have been added to the corresponding data values in the distributed data set. Due to the structural differences between DORADE and UF, these correction factors are treated differently in the two formats. For DORADE, the correction factors are recorded in the CORRECTION FACTOR DESCRIPTOR with the expectation that programs utilizing the data will apply the correction factors appropriately in the data analysis process. Therefore, the original INS attitudes are not modified except for the roll angles, which have been corrected by merging in data from the Electra's ADS data. For the UF data, all corrections have been applied because there is no place-holder for these correction factors. 11

16 3.3 Data Compression The file size for TOGA COARE ELDORA data is too large for efficient data access. The data are therefore broken into files of approximately 30 Mbytes. The data have been compressed using the data compression scheme described below: 1. Data are truncated below z = -5 km and above z = 25 km using the NOAA P-3 data compression scheme. This will reduce the data volume by about 25%. 2. To ease data handling, an artificial file mark has been inserted at the end of a sweep at about 5 minute (30 Mbytes) intervals. 4. Analysis Software This section summarizes analysis software packages which are available from NCAR for processing TOGA COARE ELDORA data. Additional analysis packages are available from Centre de Recherche en Physique de l'environnement (France), NOAA/Environmental Technology Laboratory/National Severe Storms Laboratory (Boulder), and NOAA/Hurricane Research Division (Miami). All TOGA COARE ELDORA data are available in DORADE and UF on the NCAR MSS. Additionally, DORADE and UF data are available from ATD on Exabyte tapes. The data interpolation programs (REORDER and SPRINT) have the capability to access DORADE as well as UF data. Multiple Doppler analysis and display programs (CEDRIC) make use of interpolator output and are therefore independent of radar format. 4.1 Scan-plane Data Visualization and Editing The tasks of data visualization and editing for UF data have been supported by the NCAR/ATD/RDP's PERUSAL and EDITOR. These programs continue to be available and supported, but only for UF data. A new visualization and editing software (SOLO) is being developed to replace and improve upon PERUSAL and EDITOR capabilities. The SOLO program is currently available for P-testing for interested users. A conversion routine to convert DORADE, UF, NOAA P3 data and the ELDORA FIELD format data into 'sweep files' as the input data to SOLO is available. This conversion routine will also perform aircraft ground speed removal and Bargen and Brown (1980) unfolding. In the interim users have three primary options: 1) use UF data, 2) write an interface for their own software using Fortran access subroutines which can be obtained from RSF, or 3) skip the traditional visualization/editing step and move directly into use of an 12

17 interpolator for display and analysis purposes. All three approaches are viable and are currently being used in RSF. Option 1, the use of UF data, while somewhat laborious, is certainly tractable and the necessary software packages are all available. Option 2, writing a new DORADE interface to existing software using sample programs provided by NCAR, CRPE or elsewhere, presupposes the user has alternative data visualization, editing, and analysis capabilities. This is probably the preferable choice for those users having their own personal software, and should be straightforward. Option 3, eliminating the pre-analysis visualization and editing step, while possibly a surprising suggestion, appears to be workable. (This approach has for some time been successful at NCAR/Mesoscale Microscale Meteorology Division (MMM) and Research Application Program for analysis of ground-based radar data.) Test analyses performed in RSF indicate that since much of the ELDORA data are sufficiently free from noise, simple batch editing approaches are capable of producing excellent analyses. The edit steps can occur in the radar scan space or in the process of griding, and then on the grided data. The batch editing includes elimination of data from the sea surface and below, use of NCP and local variability to eliminate residual noise, and other appropriate techniques. Our experience indicates that simple batch removal of the residual sea surface motion is also possible in this batch mode. Users of the NCAR RDSS EDITOR program are warned of a problem which was only encountered with analysis of ELDORA TOGA COARE data. Prior to TOGA COARE, the EDITOR has seldom been used for southern or eastern hemisphere (e.g. TAMEX) data analysis. An undiscovered programming error reversed the sign of the longitude and latitude for southern and eastern hemispheres. This problem has been corrected at NCAR but will affect users who obtained the software before October Contact Dick Oye or Michele Case for update information (see Table 6 for a complete listing of telephone and information). 4.2 Interpolation, Multiple-Doppler Analysis, and Display Implementation of the DORADE format data input capability for most interpolators has been accomplished at NCAR and many other institutions. Persons wishing to implement this capability on other programs can make use of sample FORTRAN or C code obtained from NCAR, CRPE, or elsewhere to ease the programming task. Our experience shows this as a straightforward programming task. At NCAR, the data interpolator REORDER (from ATD) and SPRINT (from MMM) can access UF and DORADE data. Most multiple Doppler analysis and display programs (e.g. the MMM software CEDRIC) makes use of an interpolator output format as its input. Therefore, such programs are generally not affected by the radar data format. 13

18 5. Obtaining Data, Software, and Answers To facilitate responding to requests for data, software, and assistance, a special alias (eldora@stout.atd.ucar.edu) has been established at ATD. Messages received through this address will be directed to staff who can best handle the requests contained in the message. You may send to other ELDORA staff, but please copy the message to this ELDORA address as well. All sent to "eldora" will be centrally logged (including the ATD replies). The compiled messages, with any necessary editing, will be redistributed to all interested parties. The compiled messages will also be maintained for anonymous ftp under the name ELDORA.notes, on the machine "ftp.atd.ucar.edu" ( ), in the directory -ftp/pub. It is hoped that this mechanism for compiling messages will serve to get any and all relevant information to researchers in as short a time frame as possible. Users are encouraged to obtain their own copies of the data directly from the NCAR MSS. In order for ATD to best serve ELDORA users with data from the MSS, we request an e- mail be sent to eldora@stout.atd.ucar.edu when you access the data from the MSS. The path name on the MSS is /FOFDMG/DATA/TOGACOARE/ELDORA. Individual files are named with date and time keys as well as the data format. Requests for data should indicate the required media for distribution. The preferred medium of distribution is 8 mm (Exabyte) tape (either high or low density, no internal Exabyte data compression). Due to the large data volume, only a very limited quantity of data will be provided for any user on 9-track tapes. Any such requests will be filled on a "background" priority basis because only a single 9-track drive is available. Hardware is available to facilitate batch copying of 8 mm tapes; it is likely that the quickest and easiest way to obtain data will be on 8 mm high density tapes. There is a copy charge of $25/tape for data distributed by NCAR/ATD, which covers the cost of blank tapes, shipping, and some maintenance costs of the copy equipment. The data requests should be sent to Robert Rilling at ATD. For recovery of data from the NCAR MSS, contact the Scientific Computing Division's (SCD) consulting office for techniques and details. For general questions on ELDORA data, please contact Peter Hildebrand or Wen-Chau Lee at NCAR/ATD/RSF. For the REORDER, SOLO, and the TRANSLATOR, please contact Richard Oye or Michele Case at NCAR/ATD/RDP. For the SPRINT and CEDRIC software, please contact L. Jay Miller or William Anderson at NCARIMMM. If you wish to be on the distribution list for updates concerning ELDORA, please send a short message to the address listed above; include your name, your address (internet preferred when more than one option is available), your postal address, and your phone number. 14

19 Name address Phone Number Inquiry About: William Anderson (303) SPRINT, CEDRIC Michele Case (303) Distribution on Translator, SOLO, REORDER Peter Hildebrand (303) ELDORA Data Wen-Chau Lee (303) ELDORA Data L. Jay Miller (303) SPRINT, CEDRIC SCD Consultant (303) NCAR MSS Richard Oye (303) Translator, SOLO, REORDER Robert Rilling (303) ELDORA Data Request Table 6. List of Persons to Contact Regarding ELDORA data. 15

20 Appendix A: DORADE Format Overview Al. Introduction The Common Doppler Radar Exchange Format, generally referred to as "Universal Format", (Barnes 1980) has been used extensively for the exchange of Doppler radar data. The major goal of this "Universal Format" was ease of access to ground based Doppler radar data by providing a standard for exchanging radar data. With this goal, the efficiency of the format was not a primary consideration. The tape structure was therefore designed to include all header information in every ray even though this information only changed rarely. While inefficient, this structure has allowed easy access to the data and has served the meteorology community well. A new area of scientific research opened in the early 1980s with the advent of the tailmounted Doppler radar aboard the NOAA WP-3D aircraft (Jorgensen et al. 1983; Parrish 1989). From inception the data from the NOAA P-3 airborne Doppler radars have been recorded in a NOAA field format because of the different geometry and data characteristics between airborne and ground-based radars. For subsequent analysis and combination with ground based radar data these data frequently have been translated into Universal Format. Due to the need to record navigation information unique to the moving platform, additional entries were introduced into the UF local use header. The new NCAR ELDORA airborne Doppler radar (Hildebrand et al. 1994) adds the additional complication of having two radars operating on the same platform. This capability is also emulated through beam switching or beam steering on the NOAA P-3s (Jorgensen and DuGranrut 1991). The result is that for any of these airborne radar systems, the data from two radars are recorded simultaneously. Unlike some dual radar, ground based systems which have co-located beams, these airborne radars have different beam positions. Airborne radar data are thus sufficiently complicated that Universal Format is a poor solution to the data recording needs. A new common format therefore needed to be developed. The specifications for the new DOppler RAdar Data Exchange format, DORADE, were developed by representatives from the primary producers of airborne Doppler radar data: the NCAR Atmospheric Technology Division (ATD), the NOAA AOML/Hurricane Research Division (HRD) and NOAA National Severe Storms Laboratory (NSSL), and the Centre de Recherche en Physique de l'environnement (CRPE, France). This group gathered in Miami, FL (NOAA/AOML/HRD) in April 1991 to discuss the proposed DORADE format. In addition to discussing the structure of the format, airborne radar coordinate systems and terminologies to be used in the DORADE format were defined. This group met again at the 25th Radar Conference in Paris, France, to continue the discussion on the structure and contents of the DORADE format. The initial draft of the proposed format was distributed in July 1991 to an expanded group, including scientists and programmers who will use the data, to solicit their comments. Numerous responses were received. The revised draft was then distributed in November 1991 and the most recent version of DORADE format (Lee et al. 1994b) was distributed in June

21 The DORADE format will be used to exchange data collected by the ELDORA/ASTRAIA and NOAA P-3 airborne Doppler radars and ground-based radars. Exchanged data should always be corrected as best as the facility making the tape can do, e.g. aircraft motion removed, range delay corrected, etc. A2. Design Goals for DORADE A2.1. Planning for Multiple Remote Sensing Systems The Universal Format assumed radars would be ground based radars, operated at a single PRF, with constant and uniform gate spacing and multiple variables per beam. The DORADE format is designed to meet a new, broader span of possibilities. DORADE is designed to handle a moving platform, multiple radars or instruments in the same data set, different beam positions for each radar, variable PRF, and variable gate spacing. The data structure is designed to be flexible enough to enable a reasonable lifetime of upgrades, and should not be dependent on the recording medium. Due to the scanning geometry of airborne Doppler radar, the data are collected with foreand aft-pointed radar beams which are processed as independent data sets. Whether the radar consists actually on one radar which scans fore and aft as on the NOAA P-3s (Parrish 1989, Jorgensen and DuGranrut 1991), or of two complete Doppler radar systems as in the NCAR ELDORA system (Hildebrand et al. 1994), the data are treated as if from two separate radars in the subsequent multiple Doppler analysis. On the NOAA P-3 aircraft there is also a PPI-scanning belly radar which is recorded along with the tail radar data. Other remote sensing systems are likely additions to the NOAA P-3s and the NCAR Electra. It is therefore necessary to enable the recording of data from multiple remote sensing systems, each having its own operating characteristics: scanning, range gating, data recording frequency, data types, etc. A2.2. Measurement Conventions There have been differences in the definitions of radar pointing angle, e.g. elevation, and azimuth, between the ground-based and airborne Doppler radars. The initial airborne radar systems used an elevation-over-azimuth antenna pedestal, mounted vertically on the tail of the aircraft. The "azimuth" reading from this antenna was thus the rotation angle and the "elevation" was the fore-aft tilt angle. While there is a clear logic to this installation, this application of the terms "elevation" and "azimuth" has been extremely confusing to users of the data. The DORADE specification thus includes definition of the radar beam pointing angles and all other radar data navigation parameters. 17

22 A2.3. Data Compression Due to the volume of data collected by the airborne Doppler radar systems, data compression is an important topic. The NOAA P-3 radars routinely suppress the recording of range gates with no data in order to reduce the number of tapes which must be utilized on board the aircraft. From a meteorological point of view, 30% to 70% of the radar's range bins are located below the earth's surface or far above the weather. These data can easily be deleted from the recorded data. Common forms of data compression include suppression of range gates with no data, range averaging of data at long range, and elimination of range bins below the earth surface or above some altitude. Other approaches can also be utilized. Airborne radars such as ELDORA can record at 500 Kbytes/s or more. For an 8 hour flight, about 14 Gbytes could thus be recorded. Without the use of some form of data compression this volume of data can easily overwhelm the I/O, data storage, and computational capabilities of typical workstations. Planning for flexible implementation of real-time, i.e. prior to recording, data compression is therefore a requirement for DORADE. A3. Description of the DORADE Format A3.1. Recording Media The DORADE format is not tied to a certain type of recording media, but instead should be enabled on any type of recording media. The optimal recording medium is likely to be different from the point of view of different radar operators or data analysts, and will change with time. The following rules are designed to assist with the goal of recording medium flexibility. A3.2. Overview of the DORADE Data Structure The DORADE data will be organized around "volumes", "sweeps", and "rays" of radar data. The customary high level logical construct for radar data is the volume scan. The volume scan consists of a series of radar beam sweeps which scan completely through some volume of interest. This concept works equally well for ground based or airborne radar and therefore forms the basic data block for DORADE. The sub-element of the volume scan is the radar sweep which occurs as the radar scans once between scan limits or scans through 360 in azimuth or elevation. The sub-element of the sweep is the ray which is the collection of data along one beam pointing angle, during one signal processing dwell interval. The data for each "volume" are written into one or multiple files, with a descriptive volume header written at the beginning and end of the file. The volume header includes a sensor 18

23 descriptor for each of the different radars. This volume header is followed by data from the individual radars, grouped into "sweeps". For the airborne radar the sweeps consist of complete 360 scans in antenna rotation angle. Each sweep record contains data from a complete sweep for one radar and is preceded by a sweep information block. The data from each sweep of one radar are organized by "rays", with all data for a given parameter grouped together. This logical structure is somewhat different than the field formats for the NOAA P-3 or for the NCAR ELDORA. The field formats have a version of the volume header which is somewhat larger than the normal DORADE header to include various engineering data. Thus, the ELDORA FIELD format volume headers are only written when something is changed or a maximum length of time has passed. The field headers are followed by data rays. In order to eliminate the need for sorting and buffering of data into sweeps in real time, the field format data are not organized by radar. Thus the fore and aft beams are interleaved as on the NOAA P-3 aircraft where the belly radar data are also interleaved. Consequently, any programs making use of the data must check the radar descriptors at the beginning of each ray. 19

24 Appendix B: Radar Coordinate Transformation The coordinate transformation matrix and the expression of the Doppler velocity are summarized in this appendix (For details, see Lee et al. 1994a). Following is a summary list of all symbols used in this appendix. x,y,z Leveled Cartesian coordinate systems relative to the radar u,v,w Velocity components in the Cartesian coordinate system r,k,0 Airframe relative spherical coordinate systems r Distance from the radar to a pulse volume 0 Rotation (spin) angle Elevation angle D Drift angle P Pitch angle R Roll angle H Heading T Tilt angle k Azimuth angle Vt Terminal velocity L Distance from the INS to the radar VG Aircraft horizontal ground speed WG Aircraft vertical ground speed x cos(a + R) sin H cos a sin P+ cos Hsin(0 a + R) costa + sin Hcos Psint y = r cos(oa + R) cos H costa sin P - sin H sin(o a + R) cost + cos H cos P sin sj r^ cos P c os a ( C + R)costa Ocs( + sin Psinta ) I = tan - y -1 Z 0 =sin-' - r V, = ( cos0 cos + vcoso sin ) + (w -v, - WG)sin - VGsinT, dh +L{ (1 + cos P)(cos 0 cos k cos H - cos sin k sin H)- dt dp -[sin P(cos ( cos X sin H + cos 0 sin X cos H) - sin ( cos P]-}. dt 20

25 References: Bargen, D. and Brown, 1980: Interactive radar velocity unfolding. Preprints, 19th Conf on Radar Meteorology, Miami Beach, FL, Amer. Meteor. Soc., Barnes, S. L., 1980: Report on a meeting to establish a common Doppler radar exchange format. Bull. Amer. Meteor. Soc., 61, Hildebrand, P. H., C. W. Walther, C. L. Frush, J. Testud and F. Baudin, 1994: The ELDORA/ASTRAIA airborne Doppler weather radar: Goals, design and first field tests. IEEE Proceedings, (In Press). Jorgensen, D. P., and J. D. DuGranrut, 1991: A dual-beam technique for deriving wind fields from airborne Doppler radar. Preprints, 25th Int. Conf. on Radar Meteorology, Paris, France, Amer. Meteor. Soc., Jorgensen, D. P., and P. H. Hildebrand, and C. L. Frush, 1983: Feasibility test of an airborne pulse-doppler meteorological radar. J. Climate Appl. Meteor., 22, Lee, W.-C., P. Dodge, F. D. Marks, and P. H. Hildebrand, 1994a: Mapping of airborne Doppler radar data. J. Atmos. Oceanic Technol., 11, Lee, W.-C., C. Walther, and R. Oye, 1994b: The Doppler Radar Exchange Format: DORADE. NCAR Tech. Note. NCAR/TN 403+IA. 18pp. (Also available on internet: Mosaic under Recent RSF Technical Reports). Parrish, J. R., 1989: New NOAA OAO WP-3D Doppler radar system. Preprints, 24th Conf. on Radar Meteorology, Tallahassee, FL, Amer. Meteor. Soc., Rodi, A. R., J. C. Fankhauser, and R. L. Vaughan, 1991: Use of distance-measuring equipment (DME) for correcting errors in position, velocity and wind measurements from aircraft Inertial Navigation Systems. J. Atmos. Oceanic Technol., 8, Testud, J., P. H. Hildebrand and W.-C. Lee, 1994 : A procedure to correct airborne Doppler radar data for navigation, using the echo returned from the earth's surface. J. Atmos. Oceanic Technol. (Accepted). Walther, C., and W.-C. Lee, 1994: The ELDORA field data format. NCAR Tech. Note (In preparation). 21