Freesteel Blog » Radiosondes and ranting

Radiosondes and ranting

Thursday, June 1st, 2017 at 4:06 pm Written by:

I’ve been flying around my data logger on my hang-glider and doing my own data processing with mixed results for two years now.

During this time I’ve been on the lookout for someone else’s work that I can copy.

Just yesterday I discovered the existence of Dropsondes and then Radiosondes (devices that get lifted by a weather balloon with a radio link; nothing to do with sound-waves).

The fact of their existence has been staring me in the face for years.


Those little red *S* symbols in the rasp forecast are not weather symbols for sunshine, but in fact the locations of half a dozen atmospheric “sounding” stations.

Until now I’d believed they were something involving a fancy radar beams shining up through the clouds, but it turns out it’s a freaking weather balloon with a humidity, temperature and gps sensors (they don’t bother with the barometer anymore and just use the gps altitude) that radio back data for an hour and a half till the latex balloon bursts at 25,000m and the device falls under a biodegradable parachute with a 95% of never being seen again.

The US government has a complete tour of the procedure, but the MetOffice has some automated stations which assemble and let off a new balloon every 12 hours from a robot building.

I did like their descriptions of their 1960s mechanical sonders, but lately this kind of tech has been outsourced to very boring Vaisala profit oriented corporations who invariably produce crappy closed source software that applies arbitrary filters to “clean up” the data (ie mess it about so it’s impossible to debug and fix it later when you learn what is going wrong) and are never ever ever going to invest in steerable parachute technology on the basis that they like the waste of selling to everyone cheap, crappy disposable instrument payloads if it keeps the revenue coming in at a constant rate. Who cares about progress?

These instrument payloads have a string unwinding device so that the payload is less likely to be in the wake and sun shadow of the balloon as it swings about, spinning and penduluming, and taking time to dry out after passing through a cloud.

So why the hell is it not collecting data on the way back down then, when the instrument platform is front into the airstream, it’s orientation can be controlled with the use of fins, and the circumstances are cleaner? It’s gathered the windspeeds on the way up and, without too much of a glide angle, could move to and keep itself in one vertical column or even attempt to follow back along its same track. The data sequence discrepancies would be properly reconciled by the thermal coefficients so you couldn’t get away with bodging it, and you’d get the device back on the spot which would mean it could be tested, recalibrated, made scientifically accurate and not have to be so cheap, crappy and disposable.

Heck, you could even put a valve in the balloon and reuse that so it doesn’t burst if you’re focussing on the elimination of waste. That way, and with the use of cheap plentiful hydrogen and solar panels, you could be bouncing these things up and down in the atmosphere all day, if the data was any use.

For answers to these and other questions, I’ve been reading the The WMO Intercomparison of Radiosonde Systems report from 2005 in Mauritius (no parachutes used because they fell into the sea) and the WMO Radiosonde Intercomparison report from 2010 in China (with a single paragraph mention of such a thing as a “valved balloon descent”).

They do complain about these gravity waves interfering with the balloon ascents, but that’s nothing to do with Einstein’s general theory of relativity. It’s just another data source that’s being discarded because we don’t want to make software to deal with it.

In two days of looking, I’ve not quite found what I wanted, which is some sense of how to design the software to combine and process this type of time/position series data involving environmental sensors.

But they’re still doing better than the pyrheliometer folks who rack up every few years in a comfy hotel in Davos (annual meeting place for billionaires to kid themselves that they are helping the world) and produce reports that are at least at least 50% junk — 53 pages of charts that look like exactly this:


No goddamn use even slightly.

The correct answer for handling this data and its presentation is not a printout! It is to distribute it raw in CSV files with an associated Jupyter Notebook, possibly hosted on PythonAnywhere where you could embed the statistical maths, the processing of the data and the display in useful graphs, like in this signal processing lecture notes page.

What is this?


Meanwhile, back with the “radiosonde” balloon measurements, I’ve made some progress:


Basically, you’ve got the balloon rising up at 5m/s through layers of air with quite different humidities, and high up in a near vacuum there isn’t a lot of air to affect the sensor, so it doesn’t respond very fast. (And there’s so little air, does it even matter, given that the measurements down in the cloud regions where the air actually exists is pretty dodgy.)

This is exactly the sort of thing I was experiencing in my wave flight with very sudden humidity thresholds that I presume the weather balloon data does not pick up because they’re too busy busting through and averaging to make appear clean and completely missing out on this curious phenomenon, which you’d think would have to mean something. Clouds don’t diffuse; they have hard boundaries on their tops when they have been recently produced. What’s going on there to produce these layers?

The reference is to this journal article which found a mistake in the “contamination correction” for molecules of plastic getting bound into the surface of the humidity sensor instead of water.

I still haven’t established whether they are doing anything precision-numerical with this data that would have been found out given the gross errors and dodgy data reported (though probably more reliable than fish catches), or whether it’s just being done for fun to make some pretty graphs.

Who can say?

This has taken all day. Now I really have to get on with something else. Really.

Leave a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <blockquote cite=""> <code> <em> <strong>