Hunting Hurricanes

The authors tell us about hunting hurricane using the Scanning Radar Altimeter based on the Linux system and analyzing the data with Yorick.

Figure 1. Front View of NOAA-43, One of Two WP3Ds

In March 1998, we started development of a new Linux-based data system for the NASA Goddard Space Flight Center scanning radar altimeter (SRA). The goal was to significantly reduce the system weight and volume to enable its installation on one of the NOAA hurricane hunter WP3D aircraft (see Figure 1) for the 1998 hurricane season. The SRA measures hurricane directional wave spectra and storm surge. The data will ultimately be used to help refine and improve hurricane models and improve forecasting and understanding.

The 1998 hurricane season was quite active and the SRA successfully flew in hurricanes Bonnie, Earl and Georges, collecting almost 50 hours of actual mission data.

Our principal obstacle was the short time frame until we needed to be operational onboard the hurricane hunter. The size, weight, complexity and power consumption of the SRA were also critical design items because of floor loading considerations and the limited payload capacity of the P3 aircraft when operating on long (10-hour) missions in turbulent weather conditions (hurricane-eye wall penetrations). Interrupt response time, crash-proofness and freedom from “lock-ups” were all important considerations when choosing the operating system for the SRA.

The new SRA data system, built on top of a Red Hat 4.2 system and Linux kernel 2.0.29, occupies eight inches of vertical rack space, weighs about 40 lbs, runs totally from an internal 12-volt aircraft battery and requires about 120 watts of total input power. It includes a custom ISA board with several PIC microchips which perform dedicated functions for the radar. It also includes the entire radar IF (intermediate frequency) strip, detectors and a 2ns/point waveform digitizer. No monitor or keyboard is directly connected to the SRA; instead, Linux laptops are used for all control and display. Those laptops run Red Hat 5.1 and 2.0 Linux.

The RT-Linux (Real-time Linux) software does the following:

  • Drives the waveform digitizer.

  • Computes the centroid-based range measurement between the transmit and return pulses.

  • Manages 96 automatic gain control loops.

  • Corrects for aircraft attitude and off-nadir angle.

  • Deposits formatted data in a shared memory block from which a normal Linux program extracts and records it to a disk file.

The SRA makes extensive use of Tcl/Tk and the Blt graphics library for real-time display.

Post-processing of SRA data is done with Yorick, a free and very powerful programming language that runs on Linux, a wide variety of other UNIX platforms and MS Windows.


The previous implementation of the SRA was developed in 1988 using an array of 68020s on a Multi-bus-I backplane, a CAMAC crate full of nuclear physics instrumentation and a combination of UNIX and VRTX (VRTX is a real-time kernel). VRTX ran on real-time processors and UNIX ran on the system host. The CAMAC crate was quite heavy, consumed considerable power, occupied substantial rack space and was expensive. It used hardware time-interval units (TIUS) to measure the time for a radar pulse to travel from the aircraft to the ocean and back. It used “threshold detection”, which caused the TIU to stop and a CAMAC-based waveform digitizer to acquire the return waveform. The waveform data required its own 68020 processor to “process” each waveform and extract certain data. The data were used to refine (post-flight) the range measurement made by the TIU. Threshold TIUs suffer from an effect known as “range walk”, which causes the measured range to vary as a function of the strength of the return pulse. The array of processors communicated with each other via a 4MB memory card which resided on the multi-bus. Control of the system was via a character-based terminal and real-time display was done on an SBX Matrox graphics module which was managed by its own 68020 processor. One of the 68020 processors ran UNIX; that processor ran programs which extracted radar data from the 4MB card and stored it on a 9-track magnetic tape or a disk file. The UNIX processor hosted all software development and managed the operator control terminal.

Due to its volume, weight and power consumption, we were unable to install this version of the SRA on the hurricane hunter. Limitations in the hardware signal-tracking circuits would frequently falsely trigger the system on a side lobe and effectively eliminate the true range measurement.

SRA System Description

Figure 2. Block Diagram of the SRA Sensor

The SRA is an airborne, 36GHz, down-looking, raster-scanning pulsed radar. A simple schematic block diagram of the sensor is shown in Figure 2. Its one-degree beam (two-way) is scanned across the aircraft flight track and a precise time-of-flight measurement is made for each of 64 pulses transmitted at 0.7 degree intervals across the scan. As the aircraft proceeds, a topographic image of the surface (normally ocean waves) is developed, recorded and displayed. The nominal ranging accuracy of the SRA is 10cm. Three differential carrier-phase GPS receivers are used to measure the exact location of three GPS antennas mounted in an array on top of the aircraft. A ground-reference GPS is set up where the flight originates and the ground and aircraft GPS data are processed post-flight to produce an aircraft trajectory, typically accurate to about 30cm in our application. Higher accuracies are possible when operating under less stressful flight conditions.

Figure 3. The SRA Scanner Assembly Mounted on the NOAA P3