roncli.com blog: First National Radar Test

Now playing: Namco - Sala de Masala 2000 (2:42)

I completed my first national radar test based on NWS's data. Not quite where I would like it to be, as there is ground clutter and in the northwest there are two strange looking lines that are obviously not supposed to be there... apparently NWS doesn't filter them out automatically. You can see a comparable image at approximately the same time to see the differences.

The good news is that the code obviously works. The bad news is that it took over 11 minutes to process all 136 radar sites in the continental US. NWS would have made two updates on each of those 136 radar sites in that time, not to mention that my radar data is now 11 minutes older than it was.

A good percentage of that time is spent downloading the radar data from NWS to begin with. Considering I'm on a shared T3 and it took about 3 to 4 minutes to download less than 2 MB of data, that's relatively slow.

But processing the files is taking the bulk of the time, and I may have to revisit the methods I use in order to get this to work faster. It could just be that I'm in debug mode, but I doubt that's going to help with the download speeds. Also, a lot of this is just computationally expensive. You figure there are 230 points I'm looking at for each degree around the radar - all 360 of them - for each of 136 radars. That's over 11 million pixels of data. Plus, since the map I'm displaying it on is a latitude/longitude map, I have to take into account that a nautical mile at 30 degrees of latitude is not the same as a nautical mile at 40 degrees of latitude. It gets complex, but I got a pretty good algorithm setup to do it.

So in short, there's a long way to go yet before this is practical, but the ground work and understanding is there, which is usually 80% of the work. I'll probably post more on this in the future.

Labels: Coding, NWS, Screenshot

0 Comments

Post a Comment