[ge96]

Is there any way people would be incentivized to setup a little weather station/contribute to data and get paid. Wonder if there's a model where people could make money/not game the system too. It would have to be standardized/verified to be accurate somehow.

[Aurornis]

Blockchain people have been trying variations of this for a decade. Any time you create a system that pays people for data, it will be exploited to the extreme.

I don’t think you need to incentivize people to provide weather data. Just make it easy to set up a station and get a lot of people interested. There are already hobby stations out there and networks for them.

[noosphr]

Blockchains have clearance rates that are a few dozen orders of magnitude too slow for this. Bitcoin for example clears 7 transactions per second.

[ben_w]

While that detail is true, the real problem is much more general: you have goal x, you use some proxy y for that goal, you pay people for y, they give you lots of y that may end up being the exact opposite of x.

Famously, the British found x was "fewer cobras" and y was "cobra tails", the opposite of x being "the locals bred cobras to get money for cobra tails".

Make a citizen science weather station that's free, it's all fine. Make it paid, someone's going to grab satellite pics and generate from them plausible but not necessarily accurate simulated weather station data for everywhere to get that money.

[zaphirplane]

Incentives do work in general. Sometimes they are abused. Incentives with no checks and balances are always abused. I don’t think the generalized problem you discuss above is broadly general

[ben_w]

Incentives can work, but most governments and businesses are still only mediocre at them even with enough money to throw at the problem they get to do do-overs when they get it wrong.

Trying to do this with humans on a big scale combines the worst of software development in the days of punched cards, working without anyone having given you a formal language spec, and black-hat hackers on the modern internet.

It is very very easy to pick your incentives badly; you only get feedback on a very slow cycle (in the punched card days you might run the program overnight and only find it crashed on line 32 from a typo the next morning, but it's much slower than that in meatspace); and you also need constant fine-tuning as people interested in gaming the system share their methods for doing so.

[Aurornis]

Bitcoin wasn’t designed for high throughput. I’m referring to projects like the Helium network, which rewards people for running network nodes: https://www.helium.com/

It doesn’t work as well as they wanted and it has been subject to various exploits over the years from people figuring out how to fake the participation to extract rewards.

[noosphr]

>Bitcoin wasn’t designed for high throughput

>>Commerce on the Internet has come to rely almost exclusively on financial institutions serving as trusted third parties to process electronic payments. While the system works well enough for most transactions, it still suffers from the inherent weaknesses of the trust based model. Completely non-reversible transactions are not really possible, since financial institutions cannot avoid mediating disputes. The cost of mediation increases transaction costs, limiting the minimum practical transaction size and cutting off the possibility for small casual transactions

From the bitcoin white paper.

>I’m referring to projects like the Helium network, which rewards people for running network nodes

OK, what's the clearance rate for helium?

Visa clears 35,000 a second.

[Self-Perfection]

TON blockchain showed throughput above 100k per second, about 3x more of visa. There are other blockchains with similar throughput claims. L2 networks can give even lower fees and higher throughput.

[noosphr]

Are there any good papers on it? I've found the majority of crypto literature to be between bullshit and terrible.

[squeaky-clean]

I don't think they were saying to use blockchain to do this. It's just an example that shows that if you offer financial incentives in exchange for data people will exploit it and gameify it. The reason blockchain clearance rates are so slow is because of all the effort to prevent this. You could remove PoW from bitcoin and the network would be significantly faster. It would also be dominated by people exploiting it.

It's the same thing with ad networks, most of the effort goes into verifying that an ad click was legitimate and not a bot. Or that classic story of when the British government tried to eliminate Cobras in India by paying a bounty for every dead cobra, which just led to people breeding more cobras.

[ge96]

Time series database comes in ding ding

Yeah I could see the hobby-drive there

[asteroidburger]

Many people have no problem setting up a station and giving away the data for free.

https://map.purpleair.com/

This is primarily for air quality by default, but you can get temperature, humidity, etc as well. For each station, someone paid for the hardware and is sharing the data gratis.

[pridkett]

You’ve just described Ambient Weather. What I find kinda funny about that is they still try to upsell you to get more than 1 year of data retention.

Luckily, they allow you to configure additional arbitrary locations to pump data to. I wrote a little program to drop that data into an InfluxDB database (along with PurpleAir, AirGradient, AirThings, Solar Data, and Iotawatt). The only practical use I’ve found is to look and see “When was the last time we head three days in a row that were so windy?” I suppose I could do fun stuff with Home Assistant too.

[arghwhat]

You need to collect data and run the weather models, which for good ones could require a lot of continuous compute resources.

[bigiain]

I'd be a little surprised if Google (or even Apple) haven't considered trying to use cell phone temp and pressure sensor data collected across the entire fleet of devices running their OS. Similar to the recent Android earthquake warning thing, or Google's traffic data.

Like others have pointed out though, gathering observation data is only part of the problem. Turning current and historical observations into usable and accurate forecasts is a big compute heavy task, and whoever is paying for that compute needs either government funding, which is not easy in the age of DOGE, or to charge for the forecasts.

I have a weather station that collects temp, pressure, wind speed and direction rainfall - and which has wifi and built in capability to send it's data to a bunch of web services. Sadly, it's still in the box it came in because I haven't got around to installing it and the burst of enthusiasm the inspired me to buy it has long since died. (If anyone in Sydney Australia wants it, reply here and we might be able to organise for you to come collect it.)

[eurleif]

Is there a feasible way to turn noisy cell phone temperature data into reliable weather data? Cell phones can be indoors; they can be in someone's pocket next to their body heat; they can be in direct sunlight; and they can generate a lot of heat themselves under load. And it's not just outlier phones that aren't in a position to accurately measure outdoor temperature; it's probably the majority of phones at any given time.

[bigiain]

Probably not?

But I barely understand how shit works when you operate at Google scale.

I wonder how many "Android-ish" devices like maybe in car entertainment systems are out there and reporting all their telemetry data back to Google? I wonder how much "Android" is in vehicles with AndroidAuto, and whether that hardware typically has temperature and pressure sensors like phones do?

If I had, say, a billion cars sending me data that includes temp, location, and possibly some vehicle specific CANBUS type data - I'd guess there could possibly be a signal in there that could reliably report temps and pressures at locations, based on heuristics that identify cars parked outdoors.

Same with phones. At the scale of "every single Android phone on the planet", the left over after "the majority of phone that aren't in a position to accurately measure outdoor temperature" still leaves a huge number of devices. I suspect even something stupidly simple like "What's the p99 low temperature of all the Android temp reports in a suburb?" might be a really good indicator, when "all the Android phones in a suburb" might be 10,000 devices or more?

[wmeredith]

> Wonder if there's a model where people could make money/not game the system too.

We're up against the most basic of human nature here.

[avhon1]

People pay to have a tempest weather station

https://tempest.earth/tempest-home-weather-system/

[x0x0]

That's pretty cool, thank you for sharing.

Do you have any idea if you can plug this thing into a public api to share?

[threeio]

https://community.home-assistant.io/t/weatherflow-tempest-wh...

Looks like it can be done with home assistant and then pushed up to collection locations, so I suspect things are pretty open

[paranoidrobot]

Same deal for Ecowitt: https://www.ecowitt.net/ (sign up required)

I can access any other publicly shared Ecowitt station's data.

[ge96]

interesting design, doesn't have the cliche anemometer spoons

[vlod]

>It would have to be standardized/verified to be accurate somehow.

You could do something that for the same zip/county, aggregates the results based on a certain percentage. You could weight it based on how many times a user is outside this range. (e.g. bad actors)

I just got zigbee working in my house (SONOFF Zigbee 3.0 USB Dongle Plus Gateway). Is there any recommended weather nuts out there that could recommend a weather device (that they like and is cool), just in case someone wants to create a project and is looking for data providers.

[adolph]

There are a few models for community data collection/distribution that appear reasonably successful in ADS-B and bird tracking with commercial, non-commercial and academic examples. The challenge is that there are hard costs to collecting/persisting/distributing the data which are incompatible with free (which in reality just means someone else pays).

https://www.adsbexchange.com/how-it-works/

https://www.flightaware.com/adsb/

https://www.birdweather.com/about

https://ebird.org/about

[munk-a]

This feels like a great thing for the government to do which is why NOAA/NWS have traditionally maintained these services. The data these stations produce nationally is valuable but hard to quantify on an individual station level - should the station that detects vital data about a hurricane be given a large bonus for it? If so we'll end up with extremely lopsided coverage while the information from nearby weather systems can be invaluable.

[WillAdams]

Isn't that Weather Underground's schtick?

[bigiain]

It kinda was, but they did this same rugpull and closed their free api way back before covid.

They do still claim:

"250,000+ Weather Stations

Weather Underground is a global community of people connecting data from environmental sensors like weather stations and air quality monitors "

I think it's a bit like FlightRadar24 - if you feed them your weather station (or ADBS receiver) data you get some level of free access but with non commercial use restrictions.

[WillAdams]

It's a reasonable tradeoff --- if I had a larger lot, and if there weren't already a bunch of stations in my neighborhood I'd gladly set one up.

[kawfey]

same model as flightaware et al use for crowdsourced ADS-B air traffic monitoring. You set up an ADS-B receiving station, send your feed to FA, and in return you get a premium-level account.