Hunting Hurricanes

Figure 12 is a screen shot of the SRA control laptop during hurricane Bonnie's landfall. The image on the left side of the screen is the real-time topographic display. It is gray-scale encoded so that the higher things are, the more white they appear; the lower, the darker. This image clearly shows waves on the left side of the image, the beach in the center and a very distinct dune line. We also have a color-encoded version of this program, but its interpretation is not as intuitive. The blue/brown display represents the attitude of the aircraft. It is a short Tcl/Tk script which reads aircraft attitude data captured by the SRA RT-Linux module.

Figure 12. SRA Screen in Operation during Bonnie.

The bright green display shows how we control and designate the operating conditions for the SRA. At this time, we manually find the return signal using the slider. Once found, we click the “auto” button and the system will keep the ground in the center of our digitizer window, regardless of aircraft altitude variations. The flight map is yet another short Tcl/Tk program. It extracts GPS position data from the shared memory area and uses it to map our position.

Tcl 7.6 and Tk 4.2 with Blt 2.3 are used extensively in the SRA. Initially, we thought it might be useful only for prototyping, but it soon became obvious that the X server would be the display bottleneck and not Tcl/Tk.

During development and before we purchased the laptops for control, we used a monitor connected directly to the SRA system. This meant that the X server would run there too. When we began experimenting with using a remote X server, we quickly discovered that the burden of the X server had also moved to the remote system. This was a no-effort way to automatically distribute the load across one or more computers in the system.

We wrote the image display in C using the Xview library. We used this library because we already had a book about it, and it didn't look too difficult to use. It writes each scan line directly to the display and simultaneously to a “pix-map”. When a “repaint” event occurs, the pix-map is used to repaint the whole image. A great way to put a load on the display computer X server is to grab the image map and move it around the screen. The load on the displaying computer will go through the roof, but the data system will remain unaffected.

Once we had some SRA data, we obviously needed to build some software to review it. We wanted to have SRA processing software on several machines and without licensing hassles. That way, we would be able to develop programs at home, on an office laptop (which is also used to control the SRA), on the SRA data system computer and on office Linux and Windows PCs. In total, we needed processing on at least five to ten different systems. We considered IDL, Matlab and Yorick. Our tool of choice for processing was Yorick. It is free, very powerful and will run on a wide variety of platforms including almost every UNIX machine known, Linux and Windows. It has the ability to save data so it can be read on a big-endian or little-endian machine.

Figure 13. Initial Data Product from Yorick Showing Surface Topographic Images Superimposed on NOAA Wind Plots

We first heard of Yorick from an article in Linux Journal (“The Yorick Programming Language”, Cary O'Brien, July 1998). We downloaded it and gave it a try. We like its C-like syntax and ability to load (and reload) individual functions of a program. This makes for a very powerful and flexible development environment. One of its best features is its cost—free! To put either Matlab or IDL on all the machines would have been prohibitively expensive. Since we have Yorick on the SRA data system and on the controlling laptop, we can easily analyze data in the field with minimal effort using the Linux laptop. Figure 13 shows topographic images from the SRA overlaid on a wind field plot from the August 24th flight. The sea state was above 18 meters (60 feet) on the northern flight line.

We had two or three short test flights on the NASA C-130 aircraft before we had to pack everything up and ship it to MacDill Air Force Base in Tampa, Florida, for installation on the NOAA hurricane hunter. We removed a number of bugs during these test flights, but not all. When we shipped the system, it still would not track properly.

Once we were all installed on the hurricane hunter, we had a 6-hour test flight. This permitted us to work out almost all of the bugs we had seen earlier and a few new ones. We still had a few problems with the tracking code, which would not track reliably.