New historical analysis tool in testing

[Rob H]

I've created a simple web application to view synoptic features alongside storm reports that I thought I'd share with the community. I did this partly because I wanted to play with the Google Maps API and didn't know what to plot, and partly because I have a bunch of SPC and HPC data and images and I wanted a way to mash up all the data instead of having to look at multiple images side by side or by manually querying the db.

http://66.41.225.80/storm/default.aspx
(connection might be spotty, since it's not hosted at a real location yet)

Select your filters, pick a date, and click 'Reload' to display the data. Clicking the +3/-3 and Previous/Next buttons will automatically reload. The storm reports data is from the Storm Data .csv files and covers hail, tornado, and wind reports from 1950-2009. The HPC synoptic data is from coded bulletins from roughly 2003-2010.

If you pull up 5/5/07 3z, you'll see events from 1:30z-4:30z. I did this so that you can see what the synoptic features looked like for the reports that happened in the same time frame. It didn't make sense to show tornadoes that happened 18 hours earlier when the synoptic setup looked very different.

Instead of picking a date from the calendar, or going every 3 hours, there's also an event dropdown, where you can easily skip to the next/previous outbreak, instance of massive hail, etc.

Known issues: Missing HPC data. The HPC was extremely kind to help collect and provide their synoptic data but they found some bugs with their procedures. As a result, some random chunks of data are missing, including 00z for a large number of days in the past year or two.

Future plans:

• Trying to manually add as much of the missing HPC data that I can - especially for big events
• Small rewrite to make the page more AJAXy so it doesn't reload every time a date/time is changed
• Working on getting surface and upper air data, and NEXRAD data, so you can also overlay that information
• Incorporate unofficial SPC reports and update SPC/HPC data as it comes in. Right now storm data stops after 2009, and HPC data stops sometime in Oct. 2010
• I'm playing around with some distance-calculating algorithms to see if I can sort tornadoes by type of front they were near, if they were in the warm sector or on the front, etc. We'll see how this goes, but it feels very rough
• Actually host it somewhere so people can't bombard the site and take it down easily when I'm doing other things on the box

Please let me know if you think this is useful or even mildly interesting, or if you have any bug reports or suggestions. Special thanks to Mark K. at the HPC, Nick Nolte, and a few others for inspiring me and pointing me in the right direction!

edit: Thanks Michael, it should be fixed now.

 

[Michael Towers]

Nice resource Robert, good job and thanks for sharing your efforts! While it’s not exactly a “bugâ€￾ there is an issue accurately correlating some time specific events with SPC reports. More specifically, a link to SPC from an event at 00z will produce storm reports for the following calendar day as 00z is reflected (accurately of course) as such. So for example if I’m checking event data for May 29 at 7pm the date box reflects 00z May 30 and the SPC link will subsequently correspond to May 30 reports. No big deal really though as it’s easy to navigate backward in the reports and overall I think the site will be something I find useful.

 

[Darren Addy]

A very nice app!
No event on 6-3-1980 (Grand Island "Night of the Twisters")?

 

[Rob H]

A very nice app!
No event on 6-3-1980 (Grand Island "Night of the Twisters")?

A tornado 'outbreak' is still an unofficial definition as far as I've been able to determine. The best definition I've heard is '6 per day', but that's rather difficult to handle logistically - do you display the start of the day, the first tornado, or the most prolific part of the event? I've been using 5 tornadoes in a one hour period to define an outbreak - but the problem with that is it excludes days like 6/3/1980, since there were only a few tornadoes per hour. I'll try to think of a better way to handle outbreaks. Thanks for the feedback!