Twister Tracker

[Tim Segraves]

If anyone has time I'd love some feedback on a little side project I've been working on for fun. I built a little web-app that allows you to view an estimated real-time location of a tornado based on NWS reports of location, bearing, and speed.

http://www.twistertracker.com (Home Page)

http://www.twistertracker.com/demo (Demo Warning)

I've been doing web development for about 10 years using Microsoft technologies (ASP.NET) but have recently wanted to learn Ruby on Rails so I thought this would be a good learning opportunity. The idea started out as a just a little prototype but as I continued building it out I enjoyed it so much I just kept going.

If anyone has any suggestions, ideas, criticisms, bugs, etc., I'd love to hear. I posted some more details below for anyone interested in knowing how it works in detail.

Details For Anyone Interested:

I have a cron job that runs every minute and pulls the latest national warning feed from the NWS. It parses these and stores them to a MySQL database for display on the site. If the NWS issues an update on a warning it will plot that updated location along with a new tracking line if they changed the bearing or speed of the storm. It also pushes all new tornado warnings to a twitter feed as well.

I created a mobile stylesheet so the site looks decent on an Android or iPhone. I'm toying around with the idea of some "pro" features if there is interest from people on me building this out more. I have one already where it will plot your current location (if you have GPS on your phone) on the warning details page. If anyone is going to be out chasing this next system (Sun-Tue) PM me and I can upgrade your account and you can let me know how it works.

One of my concerns with this project is I don't want the general public to get a false sense of security if the little icon has passed by them since we all know it is just an estimate and storms change direction, speed, etc. I tried to highlight this with disclaimers and text stating it is just an estimate.

So far most of my testing has been just comparing the estimated location with radar imagery and also some of the live chaser streams. I'm planning on getting some chasing in this month myself to do some more field testing.

Thanks for reading!

 

[Jason Foster]

I don't like the icon and center track idea. Why...well it's like the problem NHC has with hurricanes and tropics systems. Folks get tunnel vision on the center line, but the probably it will be that precise is low. I like the time elapse bit, and suggest rather than an icon, you could do a sweeping color change over the polygon of the warning. Perhaps keep the red transparent currently used, and uncolor/clear behind the est. tornado position but leave the outline of the warning intact.

PS (for Mark Ellinwood...that town with your namesake is in the demo video...LOL).

 

[Matt Tottle]

OMG there's a 'nader happening like 40 miles south of me and nobody told me!? Man and it's blowing right across downtown! Wait what? Oh, that was a cruel april fools joke!

It looks like a cool concept.. But I'm interested in how this program will compensate for inaccuracies in reports. It's really tough for most people to judge long distances over open land, so you could have 5 different people watching a given storm, all submitting location reports that could potentially be anywhere from hundreds of yards to miles away from the actual location (one just has to watch the SN placefile during an event to see this in action). This is going to skew your plotting and put the indicator all over the place.

 

[Tim Segraves]

I don't like the icon and center track idea. Why...well it's like the problem NHC has with hurricanes and tropics systems. Folks get tunnel vision on the center line, but the probably it will be that precise is low. I like the time elapse bit, and suggest rather than an icon, you could do a sweeping color change over the polygon of the warning. Perhaps keep the red transparent currently used, and uncolor/clear behind the est. tornado position but leave the outline of the warning intact.

PS (for Mark Ellinwood...that town with your namesake is in the demo video...LOL).

Thanks Jason. I never really thought about doing a sweeping color change of the polygon but that's a pretty good idea. One thing to note is it's not just a center line in the middle of the box. It actually draws a line from the GPS coordinates the NWS issues with the warning. I'm not exactly sure what goes in to that location but I imagine it is from the radar estimate along with spotter reports.

In terms of Ellinwood, I created the demo around my hometown of Great Bend. The line is roughly drawn over my earliest memory of a tornado. When I was in kindergarten we all had to take cover because there was one on the ground just SW of town. It lifted before it got to town but my mom later told me she saw a funnel from our house on the north side of town. I've been hooked ever since.

OMG there's a 'nader happening like 40 miles south of me and nobody told me!? Man and it's blowing right across downtown! Wait what? Oh, that was a cruel april fools joke!

It looks like a cool concept.. But I'm interested in how this program will compensate for inaccuracies in reports. It's really tough for most people to judge long distances over open land, so you could have 5 different people watching a given storm, all submitting location reports that could potentially be anywhere from hundreds of yards to miles away from the actual location (one just has to watch the SN placefile during an event to see this in action). This is going to skew your plotting and put the indicator all over the place.

Thanks Matt. I figured there would be someone from Central Kansas that got a kick out of that! Yeah, in terms of inacuracies, that's definitely going to be an issue. I'm currently just using the location coordinates the NWS issues with the warning though I have thought about trying to plot the SN reports on there as well. I think I'd have to keep using the NWS coordinates for tracking since you're right about the SN reports not being completely accurate due to what you mentioned. I could also try and use all of the reports and triangulate a position but that would be tough since the reports are all going to be off a certain amount of time from each other.

One thing to note is the NWS does a pretty good job of updating warnings from what I've seen when a storm changes direction or speed. You can see an example of this from this warning that happened in IA on 3/22/11. http://www.twistertracker.com/warnings/320 or this one from AL on 3/26/11 http://www.twistertracker.com/warnings/363 (you'll probably have to be logged in to see these)

That said I have also seen a few times where I think they goof up the report location and then issue a correction which causes the line to jump all over the place.