LCR (Loss-of-Control Risk) update

[Dan Robinson]

During the past couple of years, I have been working to develop a new winter weather parameter/scale focused on the risk of icy roads for drivers. The scale is an attempt to translate the conditions and impacts I've seen in the field over the past 2 decades into computer code.

After unsuccessful attempts to do it myself, I hired a programmer to get the concept into a working prototype. The first functioning script was released in November 2022, and I manually ran the script daily throughout that winter using NAM NEST and HRRR data. This process usually took between 30 minutes to an hour to run each day (mainly due to the size of model data downloads).

Last November (2023), I set up a server and again hired programmers to help me get an automated system going. This went online on December 4 and has been producing 6-hour LCR forecast charts every hour:

Icy Roads Forecasting and Monitoring Dashboard (LCR and BFP+ Charts)

Promoting safe driving in icy road conditions during winter weather.

icyroadsafety.com

The script produces both the LCR data and the BFP+ data set. BFP+ is a plot of all liquid-equivalent precip with different color scales for the surface temperature environment the precip is falling into.

I have designated the LCR script and process as open-source, available for anyone to download, tinker with and incorporate into their own model displays or graphics.

The parameter is a work-in-progress that I'm making updates and corrections to as each winter season progresses. The last big update was the release of version 1.2 last December that coincided with the launch of the server.

I posted this season recap video that shows the output from the LCR server through March 8.

https://twitter.com/x/status/1766571461068652886

The main site for LCR is as follows:

LCR (vehicle Loss-of-Control Risk): A new winter weather parameter and scale (v1.2.1)

Promoting safe driving in icy road conditions during winter weather.

icyroadsafety.com

That goes into the methodology of the script and has some tutorials on how to run it on your computer using Ubuntu Linux.

I would welcome any input and assistance from the weather community on improvements and further development.

 

[Dan Robinson]

All, I'm looking to do some updates to the LCR script for the upcoming season. I was considering generating Day 1/Day 2 outlooks 4 times a day, but was curious if anyone had any other ideas for something that would be more useful. The rough cost for a programmer to implement that is going to be $600-$800. As with the initial setup, I'd cover that if there is enough interest in the updates.

The URL is here: Icy Roads Forecasting and Monitoring Dashboard (LCR and BFP+ Charts)

I left the LCR server running all summer with no issues or errors encountered. It's been very stable and the cost to keep it running on AWS (an Ubuntu Linux instance) is only $20-$25 a month. Right now the server generates the maps and SFTPs the images and the updated HTML page to icyroadsafety.com which is on a separate managed hosting server. We found that there were too many security issues enabling a web server and allowing any incoming traffic to the LCR server itself.

 

[William Monfredo]

What an amazing and potentially useful project! A lot of programming's beyond me, but just a quick question...
On your LCR scale and approximate human impacts 1.2.1 (chances, likely, and serious w/colors light green through dark orange,) you mention speeds of 65, 55, and 45 mph.
For someone like me that despises ice and has had some issues with it, how does speed play into icy driving in a practical sense?

 

[Dan Robinson]

Thanks William! Loss of control is more likely at higher speeds for any given condition - the worse the conditions, the lower the speed that incidents become likely. At least that's the pattern I see in the field. The lower LCR value events tend to see only the faster drivers losing control - as the risk increases, you start seeing more incidents happening at lower speeds as well.

 

[Dan Robinson]

We have finished an update to add 24-hour charts to the LCR server output. These are generated 4 times daily at 00z, 06z, 12z and 18z alongside the hourly 6-hour charts. These are all available on the dashboard:

Icy Roads Forecasting and Monitoring Dashboard (LCR and BFP+ Charts)

Promoting safe driving in icy road conditions during winter weather.

icyroadsafety.com

We're working to add auto-posting of the 24-hour charts to the social media accounts, but that has been proving more difficult than expected. We have working scripts, but getting the necessary Python modules installed on the server has been an issue.

Regarding LCR itself, the biggest issue to tackle is the fact that models vastly over-plot light precipitation. For instance, large areas of light freezing precip are depicted around the USA every day by the HRRR. The vast majority of these do not verify. Research has shown that roads begin getting slick with as little as 0.05mm (0.002in) of QPF, and we routinely see major high-impact events with trace amounts. The "phantom icing" amounts plotted by models are mostly greater than those thresholds, which means there isn't a good way to filter them out. If the model plots any amount of supercooled droplet icing, you *have* to assign high LCR values to that unless there is a way to filter them out.

My latest attempt to filter some of those out involved using the cloud cover variable, which now limits LCR to 3 if skies are clear. That unfortunately had minimal impact on those phantom freezing rain plots.