I Got Into Weather Balloons Right Before They Stopped Flying
It started with a callsign
I sat for the FCC exam, passed the Technician element, and a few days later KO6ODW turned up in the ULS database. The plan was ordinary: a few local repeaters, maybe some digital modes, learn the etiquette.
Then I read up on radiosondes. The National Weather Service launches an instrument package over Oakland twice a day, and the telemetry is receivable with a cheap SDR while it climbs past 100,000 feet. The ordinary plan didn't last long after that.
What a radiosonde actually is
A radiosonde is a small instrument package - temperature, humidity, pressure, GPS - hanging under a latex balloon on about 80 feet of string. The NWS flies them from roughly a hundred sites at the standard synoptic times, 00Z and 12Z. On the West Coast in summer that works out to about 4 PM and 4 AM local, since the balloon goes up around an hour before the nominal time.
My local site is Oakland - KOAK, WMO station 72493. It's automated now: a Vaisala autosonde out near Doolittle Drive that inflates, tips, and releases on a schedule. Each sonde carries a ~300 mW transmitter in the 400-406 MHz band, and OAK's have historically sat close to 404.0 MHz. A flight runs about two hours, bursts somewhere above 90,000 feet, and comes down under a parachute into the East Bay hills or out toward the Valley.
From downtown San Jose that's about 35 miles southeast - too far to hear the sonde on the pad, but well within range once it climbs above the horizon. Altitude matters more than distance here.
The receiver
The software is straightforward. radiosonde_auto_rx from Project Horus does most of the work: point it at an RTL-SDR, give it your location and a frequency range, and it scans the band, identifies the sonde type, decodes the telemetry, and uploads the track to SondeHub. I had it scanning 400-406 within an afternoon.
The antenna
The SDR is the easy part; the antenna decides whether you get a clean decode or noise. I started with a whip - nothing special, just enough to find out whether the idea worked at all. It did, which was the point: get something receiving before spending money on it.
A ground-plane antenna is the sensible next step at 404 MHz - a quarter wave up there is only about 18 cm, so the whole thing stays compact - and I've got one arriving tomorrow. The timing, as it turns out, is not ideal. More on that below.
First decodes
The whip was enough to get started. A little after 4 PM the first signal came up out of the noise, auto_rx switched from scanning to a serial number, and the dashboard filled in: pressure dropping, altitude climbing about a thousand feet a minute, and a track moving across the SondeHub map with my station listed as a receiver.
I caught a couple of clean flights that way. One climbed through the usual California subsidence inversion - warm, dry air aloft, the cap that keeps our marine layer in place - which showed up clearly on the SkewT. After burst I watched the landing predictor shift around and started thinking about actually recovering one. That's the natural next goal: not just decoding the sonde, but going out and finding it. I was close to trying.
Then the launches stopped
The first miss I wrote off as a scrub - high winds or a passing fault, it happens. But there was no morning flight either, and none the next day, or the day after.
So I pulled the launch history. Oakland flew 58 of a possible 60 soundings in June - both cycles, essentially the whole month, right up to the 30th. In July: none. A clean stop at the month boundary, both cycles at once.
That doesn't look like weather, and it doesn't look like a random hardware failure - neither one lines up with the calendar. I checked for an official notice and found nothing naming OAK: no service-change notice or public information statement. The context is easier to find. The upper-air network has been stretched all year; federal staffing cuts have led the NWS to halt or reduce launches at around a dozen sites, morning flights across the Lower 48 are down by roughly half, and the effect on forecast quality has been building for months.
Oakland is a slightly odd case against that trend. The common pattern is a morning launch being dropped or moved to the afternoon. OAK was flying both cycles reliably in June and then went dark entirely on July 1, which points to a discrete cause - a suspension effective the first of the month, or the autolauncher or its gas supply reaching a limit - on a network without much slack to fix it quickly.
Where it stands
That's the current state: ticket, receiver, whip antenna, a few good decodes with my callsign on the map - and then the thing I'd set it all up for went quiet. The ground plane still arrives tomorrow, which will make it the best-equipped station on the block for receiving nothing.
The dongle is still running and auto_rx is still scanning 400-406. When Oakland launches again, I'll be listening. If you're near the Bay and have an RTL-SDR sitting in a drawer, it's not a bad time to get it set up, so it's ready when the launches come back.
- Mark, KO6ODW