Full Disk Images of Earth from GOES-17 | Hacker News Reader

Full Disk Images of Earth from GOES-17

esorensen.com

218 pointsJerry259 comments

This kind of thing could replace some uses of blockchain. The pattern of clouds on earth every day is unique, so anything that needs a timestamp with 1-day resolution could just tie it to a hash of today's cloud image pixels gathered from many receivers around the world, each taken at noon from the POV of a given geostationary satellite. That way one could prove that an event took place no earlier than today, especially if you hash each cloud photo with the hash of the one from the day before, which creates a cloudchain. The chain's ordering is verifiable from multiple sources and created wholly external from any human agency or pseudo-random number generator.

Speaking of which, you could also use the pixels (or some permutation of them) as a random number seed that changed every day.

So essentially instead of hash-based computer proof of work, you're suggesting looking at cloud patterns (say, as the sun's proof of work)? Sounds like a great way to save energy, if we can ensure image authenticity/integrity.

The vulnerability is the one satellite everybody is looking at. But that could be mitigated by having a network of ground stations looking up and asking "is it cloudy overhead?" With enough of these sample points scattered around the planet and correlated with the satellite's imagery, one could make hacking the satellite impractical.

Why not taking the clouds of Jupiter? Then you could use telescopes and multiple satellites alike because that far away parallax shouldn’t change a lot.

Jupiter's not always visible from Earth (sometimes it's behind the Sun). Earth is always visible from a geo sat (provided the satellite is working).

Let's use the sun then :)

> you could also use the pixels (or some permutation of them) as a random number seed that changed every day

Except that would make your PRNG easy to guess, since this data is publicly available to everyone?

It wouldn't be useful for crypto (except maybe as a salt?) but it might be useful for repeatable random simulations.

Repeatable random simlations that cannot for some reason just check the date and use the date (and some salt) for the RND seed? What simulations are we talking about here? Must be some obscure things?

Yeah I'm still trying to think of a good example for that.

That's the point, it proves the day. I think there's still a challenge to solve which is mapping the points from which to construct the hash, because you need a low enough quality to agree on pixels but a high enough quality to get enough entropy to be strong as an time consensus authentication source.

Is it any better than any other source of random or extremely difficult to predict data? Seems like you could just use data from the stock market or some radio antennas or telescopes.

The stock market is human-generated so somebody with enough money could nudge it in one direction or another. Radio antennas or telescopes could work too; I'm just not aware of anything they see that meets some definition of "random" except for maybe the surface of the sun, and the pattern of clouds on the earth.

How about clouds on another planet? Makes manipulation even harder and easier to observe from earth itself

It's actually quite difficult to observe clouds on other planets from earth (or earth orbit) due to resolution limits of the combination of reasonably-sized telescopes and usable wavelengths.

Venusian clouds are easy to see with earth-based telescopes. Problems: There's not a lot of variation -- it's solid cloud cover. And Venus has phases and we never see it as a 100% disc (the Sun is in the way).

> with enough money could nudge it in one direction

can only do for a limited period of time, no one can control the graph tick-by-tick for the whole day.

> Radio antennas or telescopes could work too

Antennas are vulnerable to noise, anyone with enough effective radiated power placed near enough to it can nudge it one direction or another.

> except for maybe the surface of the sun

Perhaps, but overall the Sun follows 22-year cycles. [1]

[1]: https://en.wikipedia.org/wiki/Solar_cycle

In this case you have to trust whatever oracle you are using. The images need to be verbatim so everyone would have to use the same centralized source, which means it is not the same as a more decentralized source.

Sound or tv -card should generate white noise.

What other hard to predict data are you thinking?

The cloud data is independently verifiable and somewhat globally accessible, which is the main benefit. In a court of law you can refer back to any number of facts to corroborate the data.

But yes you could use stock prices or sports scores to the same effect.

To your point of randomness, I always thought the cloudflare lava lamps were pretty interesting:

https://blog.cloudflare.com/lavarand-in-production-the-nitty...

Pressaged by SGI's LavaRnd:

https://web.archive.org/web/19971210213248/http://lavarand.s...

https://en.wikipedia.org/wiki/Lavarand

https://www.wired.com/2003/08/random/

Truly nothing new under the sun. I salute you, thx for sharing.

Coincidentally I used weather information to generate a random number, I used {{PollenCount}}{{LowTempCelsius}}{{HighTempCelsius}}{{WindSpeedKph}}{{Humidity}}{{UvIndex}}{{SunriseAt}}

I know this would collide eventually but hey its just a hack for my IFTTT applet [1]

1. https://imgur.com/y0PPH7E

Just add a monotonically increasing number (like milliseconds from the big bang) that will eliminate the collision, but not add much entropy.

True in general. But this IFTTT platform has some limited supply of types of placeholders, we are not allowed to insert arbitrary numbers.

Just use any CCD with the gain turned up. People have been doing this for a long time :)

http://www.lavarnd.org/lavarnd.html

Blockchain timestamping are often more concerned with establishing that things happened no later than today.

The bitcoin genesis block (https://en.bitcoin.it/wiki/Genesis_block) included a quote from the Times.

    The Times 03/Jan/2009 Chancellor on brink of second bailout for banks

You mean no earlier.

The Times quote - no earlier.

Publicly verifiable blockchain embedded message - no later.

Exactly what I meant. Proving the 'no earlier' part is relatively easy and low-tech. The 'no later' part is what requires the blockchain.

I agree. It wasn't obvious from the context, since you provided newspaper title from the genesis block as an example.

> That way one could prove that an event took place no earlier than today

I probably miss something, what prevent you from taking a old event and tag it with today hash ?

It would only work where there was no motivation to show an event happened earlier than it did, but rather as late as possible. Such as the proverbial kidnap photo where the victim is holding a copy of today's newspaper.

If someone else also tagged the event, you can’t.

Do things this way creates a single point of failure (the weather satellite), which is one of the thing cryptocurrencies try to avoid.

Looks like the site is down due to an HN hug of death.

If you do get through, this is much more interesting than the title makes it sound. It's not just the images, but how you can get them directly from the satellite with your own radio and antenna.

Is there a word for the converse of "clickbait"?

Clickwait

> Is there a word for the converse of "clickbait"?

The opposite of bait is repellent, so in that sense it would be "clickrepellent". However, I don't think that gets across the point that it is the inverse of "low value content with a misleadingly appealing headline", i.e. it is high value content with a headline that doesn't adequately reflect the value of the content. For this I was going to suggest "clickfish", as in the valuable objective of bait (assuming a fishing metaphor). However, a quick internet search reveals this is a widely used term already, e.g. for an internet surfer who falls for (and reposts) clickbait, or the name of anti-clickbait software. Then I thought of abstracting this a level with "clicksupper", as in fish supper, a wholesome meal made from fish and served with chips. Or how about "clicktreasure", thinking of buried treasure. But then again, how about simply "anti-clickbait"?

Clickwin

Here’s an archived version of the page: http://archive.is/GIhT3

And the HD video he mentions is on:

https://www.youtube.com/watch?v=qGS86jIHusE

It's also possible to receive infrared images using the same technique, which can be used in part to visualize the water vapor content of the atmosphere:

https://www.goes-r.gov/education/ABI-bands-quick-info.html

Some parts of the imaging instrument (ABI) on GOES-17 responsible for keeping it cool failed shortly after launch, causing the temperature to vary over the course of the day:

https://www.goes-r.gov/users/GOES-17-ABI-Performance.html

This doesn't affect the "visible" light images too much, but can cause all sorts of problems for the infrared, which is used scientifically for various cloud measurements, sea surface temperature, land surface temperature, etc. It's still up in the air how useful we can make this data.

I don't quite understand it from skimming the article. How are you able to talk to the satellite on the protocol level? Is it continuously sending down the bits in a known format? Do you have to authenticate with it?

In this particular case, the author is using an SDR device to caputure data emitted by the satellite, and then decode it using software (goestools). There is no interaction with the satellite, it's just a device listening to a broadacst emission.

As mentioned in the article, there are many interesting related projects:

* https://www.rtl-sdr.com/rtl-sdr-tutorial-receiving-noaa-weat...

* https://hackaday.com/2013/01/12/listening-in-on-weather-ball...

* http://www.sdr-satellites.com/Weather

This is really cool! I keep meaning to try something like this just for the thrill of receiving the data myself.

Though if receiving the data with your own antenna and SDR isn’t exciting, and you just want to use the data, AWS hosts these (GOES-16 and 17) data on S3 for everyone to use: https://registry.opendata.aws/noaa-goes/

(Disclaimer: I work on the team that helps host these datasets)

I did wonder what full disk images of Earth might contain, considering various block formats to contain various data collected from the globe. Then I read the article.

I assume it’s a similar process to take the data from the European EUMETSAT satellites?

Other geostationary satellites are available:

https://en.wikipedia.org/wiki/Weather_satellite#Geostationar...

You quite often see a bank of clocks for each time zone, it would be good to have a geostationary image like this, from several different weather satellites.

The Meteosats don't directly disseminate, there is a service, EUMETCast, which broadcasts from commercial communications satellites, although you need a decryption key: https://www.eumetsat.int/website/home/Data/DataDelivery/EUME...

The rationale behind this was that, in this way, you can receive the stream with off-the-shelf satellite TV hardware and a PC, and without special radio operator knowledge. This was of course well before the advent of cheap SDR devices.

Nice!

A bit unrelated but If you haven't seen it be sure to check out NASA's EPIC[0] aboard their DSCOVR[1] satellite

[0] https://epic.gsfc.nasa.gov/ [1] https://www.nesdis.noaa.gov/content/dscovr-deep-space-climat...

Another way to pull them: http://rammb-slider.cira.colostate.edu/?sat=goes-17

Nice. Makes me wonder what the hell else that's useful could be slurped from up there.. Must be hundreds of feeds to listen to.

Site's down. Hackaday has a writeup: https://hackaday.com/2019/05/03/__trashed-28/

And Sorensen has been active on Reddit with this: https://www.reddit.com/r/RTLSDR/comments/bj61f9/24_hour_time...

Pretty cool. Anyone has any idea how the copyright situation looks like for these images?

Most US Government produced works are public domain.

https://www.ospo.noaa.gov/Organization/About/use.html

Very nice information ! geo stationary satellite 17 (GOES-17) would really help in forcasting weather and taking full span images of earth

Reading the headline I thought HN overnight settled on flat-earthers being right and this link provides uncensored images for the first time.

Reply on news.ycombinator.com