3 ms·

If only there were someone working on high-accuracy weather forecast systems for convective systems using existing cheap networked sensors ;). I would love to k

by cryptoz 12y ago

If only there were someone working on high-accuracy weather forecast systems for convective systems using existing cheap networked sensors ;). I would love to know how much money I could save SpaceX by providing this forecast more accurately with much more lead time. What is the cost to scrub a launch due to the weather at the last minute?

ceejayoz 12y ago
A forecast doesn't stop a popup thunderstorm from entering the no-go zone.
- cryptoz 12y ago
  Of course not. But it would help SpaceX plan and prepare much more in advance than 3 minutes. There is certainly a cost associated with getting everything set up, and then postponing the launch. Could they have planned for this earlier and not got everything ready?
  - ceejayoz 12y ago
    > Could they have planned for this earlier and not got everything ready? Probably not. They were a go up until just three minutes before launch. I doubt any extant or near-future forecasting system is able to tell you that a specific storm cell will be in a specific spot at an exact minute all that far in advance.
    - cryptoz 12y ago
      > I doubt any extant or near-future forecasting system is able to tell you that a specific storm cell will be in a specific spot at an exact minute all that far in advance. Well, this is exactly my plan, and I'm going to do it ASAP. Prototype should be done by summer. There are millions of little weather stations all over Florida that nobody is using (smartphones). By connecting those stations together and pulling the live sensor data into forecasts, I do in fact believe that these small thunderstorm cells should be very well-forecasted if we take in this dense dataset. This forecast system doesn't exist yet, you're right. But it should exist in the next couple months/years, and when it does, it should be able to provide this kind of forecast. BUT only if that is worth money to a group like SpaceX - is it? That's really my question. I know I can build the technology, but will anyone like SpaceX buy, and at what price?
      - MichaelGG 12y ago
        What do cell phones have that are particularly useful to detect lightning? Or look at other sources for lightning, then thousands of barometers to predict movement?
        ceejayoz 12y ago
        Cell phones aren't needed to detect lightning. It's the pressure measurements that are potentially valuable in making a more accurate map than can be gathered by scattered weather stations. http://www.lightningmaps.org/realtime http://www.lightningmaps.org/realtime shows lightning in realtime as they're detected by stations across the planet. If you turn on the "stations" display, you can see detectors in Utah spotting lightning in the Gulf of Mexico.
      - mikeash 12y ago
        How accurate would it be? Even if you can predict with, say, 95% accuracy that a thunderstorm will cause a launch scrub, SpaceX might want to give it a try anyway in case that 5% comes true. They want to launch ASAP, and they'd look dumb if they called it off only to find that the weather was fine. Accurate forecasts are no doubt extremely valuable in many cases, but for this specific case I don't see it changing much.
    - th0br0 12y ago
      To cryptoz (can't reply, probably too deep?): I was going to argue that it would be terrible difficult to build up that smartphone network. But then I saw that you do PressureNet (awesome tool btw!) and, oh well... looks like you've definitely got the network part done already. OTOH, why not just ping @elonmusk on twitter?
      - cryptoz 12y ago
        Well thanks! But actually, building the network is indeed a big challenge. We've spent the better part of our 3 years just thinking and implementing methods to do it. We've arrived at an SDK model [1] where we work with app partners to integrate simple sensor code and contribute to our network. So we're able to grow quite quickly compared to a single app to download, but it's still not easy getting the necessary scale. We're aiming for about 10 million sensors, currently have 0.5 million. So we're getting there. I've sent @elonmusk a few messages over the years but no replies so far. I am @vallesmarineris, so I think one day he'll notice that I already live on Mars and we'll chat. [1]: https://github.com/Cbsoftware/pressureNET-SDK https://github.com/Cbsoftware/pressureNET-SDK
      - sandstrom 12y ago
        preassurenet is cool! It always reminds me about this interchange: https://news.ycombinator.com/item?id=8292091 https://news.ycombinator.com/item?id=8292091 Probably easier to google up someone on the launch team and ask them.
  - barkingcat 12y ago
    I think you're thinking about the problem the wrong way. There is a cost to postponing the launch, but compared to having their rocket blown up by lightning on launch, it's nothing. As far as I know, the budget for these launches already includes x number of retries - since scrubbing launches and retrying in another window is actually standard procedure - so they are not "loosing" as much money as you think. It's only if they exceed the # of retry attempts that they've budgeted for that a solution like yours becomes viable, and even then, you're not fighting against "the cost of getting things set up". You're fighting against the cost of having their rocket blown up on launch because of bad or unforseen weather events. That's why they scrub - not because the weather is too bad to fly - it's because there could be an unforseen event happening because an anvil cloud (which has lightning in it) - merges with another large storm formation at 10 nm - basically, it's insurance against the rocket blowing up. If you can insure your weather detection solution in the high hundreds of millions for a positive launch, then there would be a market for it. And to be successful, your product will need to cut that 3 min downwards, not upwards. So that mission control can have a better confidence of scrubbing a launch, say within 10-s of launch time vs 3 min. You get what I mean? Better weather detection allows mission control to delay longer until the best guess estimate of when to hit the STOP button, not so that they can plan ahead to put a launch at xyz date/time because a weather prediction system says so.
Animats 12y ago
There's plenty of weather data for that area.[1] That's the world's most heavily instrumented area for weather.[2] A huge amount of effort goes into predicting lightning hazards. There are electrostatic field measuring devices for miles around to give warning of pre-lightning conditions. Right now, there is a major thunderstorm system north of the launch area, and a smaller one south of it. They were trying to launch through the gap. [1] http://www.accuweather.com/en/us/cape-canaveral-fl/32920/weather-radar/2230906 http://www.accuweather.com/en/us/cape-canaveral-fl/32920/wea... [2] http://commons.erau.edu/cgi/viewcontent.cgi?filename=4&article=1027&context=stm&type=additional http://commons.erau.edu/cgi/viewcontent.cgi?filename=4&artic...
andor436 12y ago
How are you going to sense thunderstorms with a smartphone? Temperature / Barometric pressure changes? User input? Weather Underground has been doing the latter for a couple of years, although it doesn't seem to be altering their "current conditions" reporting yet: http://www.wunderground.com/about/pr/news.asp?date=20131121 http://www.wunderground.com/about/pr/news.asp?date=20131121 It's a great idea - I think there are a bunch of similar applications that could leverage phones like this.
- cryptoz 12y ago
  Thanks - I have about 400,000 barometers connected globally right now. Our data archive is about 1.5B measurements, we're getting about 5M per day. We will use this as the primary new input to WRF model runs, outputting high-res forecasts. The network I run is called PressureNet, you can see our website at http://pressurenet.io http://pressurenet.io or have a look at our flagship Android app: https://play.google.com/store/apps/details?id=ca.cumulonimbus.barometernetwork https://play.google.com/store/apps/details?id=ca.cumulonimbu... Early initial targets are large convective systems (like these) that are difficult to predict. Our dense barometer network should resolve them far better than other existing networks. We also do Current Conditions like WU, but that idea needs a lot of work before anyone makes a valuable product on it alone. Typically those readings are used to validate and tweak models, not make forecasts. Other sensors that will prove useful: cell signal strength, thermometer, humidity, UV sensor, ambient light sensor, etc. And yes, there are lots of other uses! Like earthquake mapping and alerting, I wrote a blog post mentioning some ideas and the importance for the code to be open source: http://www.jacobsheehy.com/2014/02/open-source-software-is-important-for-modern-science/ http://www.jacobsheehy.com/2014/02/open-source-software-is-i...

by cryptoz 12y ago

ceejayoz 12y ago

A forecast doesn't stop a popup thunderstorm from entering the no-go zone.

cryptoz 12y ago
Of course not. But it would help SpaceX plan and prepare much more in advance than 3 minutes. There is certainly a cost associated with getting everything set up, and then postponing the launch. Could they have planned for this earlier and not got everything ready?
- ceejayoz 12y ago
  > Could they have planned for this earlier and not got everything ready? Probably not. They were a go up until just three minutes before launch. I doubt any extant or near-future forecasting system is able to tell you that a specific storm cell will be in a specific spot at an exact minute all that far in advance.
  - cryptoz 12y ago
    > I doubt any extant or near-future forecasting system is able to tell you that a specific storm cell will be in a specific spot at an exact minute all that far in advance. Well, this is exactly my plan, and I'm going to do it ASAP. Prototype should be done by summer. There are millions of little weather stations all over Florida that nobody is using (smartphones). By connecting those stations together and pulling the live sensor data into forecasts, I do in fact believe that these small thunderstorm cells should be very well-forecasted if we take in this dense dataset. This forecast system doesn't exist yet, you're right. But it should exist in the next couple months/years, and when it does, it should be able to provide this kind of forecast. BUT only if that is worth money to a group like SpaceX - is it? That's really my question. I know I can build the technology, but will anyone like SpaceX buy, and at what price?
    - MichaelGG 12y ago
      What do cell phones have that are particularly useful to detect lightning? Or look at other sources for lightning, then thousands of barometers to predict movement?
      - ceejayoz 12y ago
        Cell phones aren't needed to detect lightning. It's the pressure measurements that are potentially valuable in making a more accurate map than can be gathered by scattered weather stations. http://www.lightningmaps.org/realtime http://www.lightningmaps.org/realtime shows lightning in realtime as they're detected by stations across the planet. If you turn on the "stations" display, you can see detectors in Utah spotting lightning in the Gulf of Mexico.
    - mikeash 12y ago
      How accurate would it be? Even if you can predict with, say, 95% accuracy that a thunderstorm will cause a launch scrub, SpaceX might want to give it a try anyway in case that 5% comes true. They want to launch ASAP, and they'd look dumb if they called it off only to find that the weather was fine. Accurate forecasts are no doubt extremely valuable in many cases, but for this specific case I don't see it changing much.
  - th0br0 12y ago
    To cryptoz (can't reply, probably too deep?): I was going to argue that it would be terrible difficult to build up that smartphone network. But then I saw that you do PressureNet (awesome tool btw!) and, oh well... looks like you've definitely got the network part done already. OTOH, why not just ping @elonmusk on twitter?
    - cryptoz 12y ago
      Well thanks! But actually, building the network is indeed a big challenge. We've spent the better part of our 3 years just thinking and implementing methods to do it. We've arrived at an SDK model [1] where we work with app partners to integrate simple sensor code and contribute to our network. So we're able to grow quite quickly compared to a single app to download, but it's still not easy getting the necessary scale. We're aiming for about 10 million sensors, currently have 0.5 million. So we're getting there. I've sent @elonmusk a few messages over the years but no replies so far. I am @vallesmarineris, so I think one day he'll notice that I already live on Mars and we'll chat. [1]: https://github.com/Cbsoftware/pressureNET-SDK https://github.com/Cbsoftware/pressureNET-SDK
    - sandstrom 12y ago
      preassurenet is cool! It always reminds me about this interchange: https://news.ycombinator.com/item?id=8292091 https://news.ycombinator.com/item?id=8292091 Probably easier to google up someone on the launch team and ask them.
- barkingcat 12y ago
  I think you're thinking about the problem the wrong way. There is a cost to postponing the launch, but compared to having their rocket blown up by lightning on launch, it's nothing. As far as I know, the budget for these launches already includes x number of retries - since scrubbing launches and retrying in another window is actually standard procedure - so they are not "loosing" as much money as you think. It's only if they exceed the # of retry attempts that they've budgeted for that a solution like yours becomes viable, and even then, you're not fighting against "the cost of getting things set up". You're fighting against the cost of having their rocket blown up on launch because of bad or unforseen weather events. That's why they scrub - not because the weather is too bad to fly - it's because there could be an unforseen event happening because an anvil cloud (which has lightning in it) - merges with another large storm formation at 10 nm - basically, it's insurance against the rocket blowing up. If you can insure your weather detection solution in the high hundreds of millions for a positive launch, then there would be a market for it. And to be successful, your product will need to cut that 3 min downwards, not upwards. So that mission control can have a better confidence of scrubbing a launch, say within 10-s of launch time vs 3 min. You get what I mean? Better weather detection allows mission control to delay longer until the best guess estimate of when to hit the STOP button, not so that they can plan ahead to put a launch at xyz date/time because a weather prediction system says so.

Animats 12y ago

There's plenty of weather data for that area.[1] That's the world's most heavily instrumented area for weather.[2] A huge amount of effort goes into predicting lightning hazards. There are electrostatic field measuring devices for miles around to give warning of pre-lightning conditions. Right now, there is a major thunderstorm system north of the launch area, and a smaller one south of it. They were trying to launch through the gap. [1] http://www.accuweather.com/en/us/cape-canaveral-fl/32920/weather-radar/2230906 http://www.accuweather.com/en/us/cape-canaveral-fl/32920/wea... [2] http://commons.erau.edu/cgi/viewcontent.cgi?filename=4&article=1027&context=stm&type=additional http://commons.erau.edu/cgi/viewcontent.cgi?filename=4&artic...

andor436 12y ago

How are you going to sense thunderstorms with a smartphone? Temperature / Barometric pressure changes? User input? Weather Underground has been doing the latter for a couple of years, although it doesn't seem to be altering their "current conditions" reporting yet: http://www.wunderground.com/about/pr/news.asp?date=20131121 http://www.wunderground.com/about/pr/news.asp?date=20131121 It's a great idea - I think there are a bunch of similar applications that could leverage phones like this.

cryptoz 12y ago
Thanks - I have about 400,000 barometers connected globally right now. Our data archive is about 1.5B measurements, we're getting about 5M per day. We will use this as the primary new input to WRF model runs, outputting high-res forecasts. The network I run is called PressureNet, you can see our website at http://pressurenet.io http://pressurenet.io or have a look at our flagship Android app: https://play.google.com/store/apps/details?id=ca.cumulonimbus.barometernetwork https://play.google.com/store/apps/details?id=ca.cumulonimbu... Early initial targets are large convective systems (like these) that are difficult to predict. Our dense barometer network should resolve them far better than other existing networks. We also do Current Conditions like WU, but that idea needs a lot of work before anyone makes a valuable product on it alone. Typically those readings are used to validate and tweak models, not make forecasts. Other sensors that will prove useful: cell signal strength, thermometer, humidity, UV sensor, ambient light sensor, etc. And yes, there are lots of other uses! Like earthquake mapping and alerting, I wrote a blog post mentioning some ideas and the importance for the code to be open source: http://www.jacobsheehy.com/2014/02/open-source-software-is-important-for-modern-science/ http://www.jacobsheehy.com/2014/02/open-source-software-is-i...