New Orleans Tornado Analysis in Real-Time

Matt Gaffner
Feb 10, 2017 · 4 min read

As most everyone knows, “Tornado Season” correlates with spring and summer. That’s typically when we see the most tornadoes. However, that doesn’t mean that tornadoes won’t happen any other time of the year. They’re just less likely.

That being said, there were several rare, but not improbable tornadoes that struck parts of Louisiana, Mississippi, Tennessee, Georgia, and Florida on February 7, 2017 as part of a larger storm system that swept across the Southeast.

Reported Tornadoes — 2/7/2017

If you happen to follow me on Twitter (@mattgaffner), I retweeted an awesome map comparison from a coworker, Andrew Thut (@MNWeatherman), showing power outages compared to a TornadoTrax product that we create at Weather Decision Technologies.

Given the correlation, I thought it might be cool to show how the real-time TornadoTrax polygons lined up with Local Storm Reports (LSRs) and the actual tornado damage path as surveyed by the National Weather Service (NWS). Additionally, I’ll point out a few trends that I’ve noticed.

Let’s start by looking at the LSRs which are readily available and tend to be reported in near real-time. They’re the first go-to for tornado data for a lot of people. The orange triangles are the point locations of the reported tornadoes as recorded by the NWS. The red triangles are the point locations I have adjusted by moving the location of the tornado reports to the intersection referenced in the comments documented by the NWS.

Uncorrected vs. Corrected LSRs

Often times LSRs will contain specific information, like an intersection, in the comments/notes. Sometimes more than one intersection are included in those notes. Two simple dots on a map essentially turn into 4 if you take the time to manually plot the specific places that are noted. Keep an eye on the two additional red dots that weren’t noted in orange. They’ll correlate almost directly with the damage path. Even knowing that, using only the tornado reports would make it very difficult to approximate the area where the tornado was actually on the ground.

Don’t think for a minute that I’m trying to imply that LSRs are worthless when assessing tornado damage in real-time. Alone, they’re usefulness is limited, but they’re actually quite useful in conjunction with TornadoTrax. Because our TornadoTrax algorithm is radar-based, it can only identify if a storm is rotating above the ground. The algorithm can’t definitively determine if a tornado is on the ground. In fact, the number of rotating storms that actually produce a tornado make up only a small percentage of rotating storms. In reality the tornado LSRs actually serve as real-time ground truth to the TornadoTrax swath.

Seeing tornado reports coincident with TornadoTrax (or even near by) is a strong indicator that there has been a tornado on the ground within that blue swath. Together, TornadoTrax and tornado reports quickly help hone in on the area that was likely impacted by a tornado in real-time. Note that the LSR north of the blue swath still helps validate TornadoTrax due to it’s proximity to the path, but the damage is still likely within the swath. LSRs aren’t always precise enough to define a damage path based on how the are documented.

Fast forward a day or two and the NWS will typically put out a damage survey that defines the actual damage path of a tornado, especially if it moves through a populated area.

LSRs, TornadoTrax, and NWS Damage Survey

The damage survey lines up pretty well with the strongest areas of rotation (darker blues). The two separate damage paths fall close to the local maxima. I have typically found that the actual damage path tends to fall within the TornadoTrax swath and lies slightly to the right of the center of the swath relative to the direction the storm is moving.

NWS Damage Survey overlaid on WDT Real-Time TornadoTrax

The EF-0 rated part of the path also extends beyond TornadoTrax. We don’t always see this. When we do it’s as the tornado is weakening, and the damage tends to be EF-1 or lower as in this case.

It’s also worth noting that their are two tornado LSRs (larger red triangles) that lie directly within the NWS’s damage path. These are the two LSRs that I manually moved to the exact intersection noted in the comments. Had they not been moved, they would have been stacked on top of the LSR that lies north of the actual tornado path.

So, whether looking to restore power, route first responders, estimate insurance claims, or determine if your companies assets may have been impacted, combining LSRs and TornadoTrax is the best way to assess the possible impact in real-time. If you’re a GIS pro, we’ve published both TornadoTrax and LSRs as real-time updating, fully attributed map services that you can access directly using the REST endpoints.

Care to check out the data yourself? Let me know. We can easily provide the data in GeoJSON, KML, or shapefile. We can also provide access to our ArcGIS Server REST endpoints for a demo.

Thanks to Steve Miller.

Matt Gaffner

Written by

Weather Nerd. GIS Geek. Analyzing all things spatiotemporal. DTN Weather — matt.gaffner {@}dtn.com