[EAS] USVI and PR: polygons versus political boundaries

[Sean Donelan]

On Thu, 7 Sep 2017, Botterell, Arthur at CalOES wrote:
> Not sure I understand, Sean.  Did you generate a plot of the data I 
> posted?  If so, please share it.  I can't take any responsibility for 
> the polys you posted earlier... do you know where they came from 
> originally?

That's why a transport protocol standard is not really a good way to list 
requirements or a concept of operations.  The polygons in National Weather 
Service warnings are generated by the National Weather Service GIS 
systems.  NWS GIS systems use a a maximum of 20 vertices, and two digits 
after the decimal point, or roughly 1.1km (about 3,600 feet compared to 
NYC's requirement for 568 feet).

As NWS polygons are simplified from its GIS systems, there are often 
overlapping areas and gaps between areas, especially with warnings based 
on civil jurisdictions such as states, counties and islands.

GIS systems that generate 15 digits after the decimal point tend to the 
result of bad software programming rather than better precision. Yes, 
more sample points provide higher resolution, but also take longer to 
transmit.  The U.S. Geological Survey has extremely accurate maps of the 
entire United States with resolution of 1 meter or better. But high 
resolution GIS plots are huge.

As you point out with your warn.pbs.org example, its very easy for a
computer to use a lookup table to identify an area.  On the other hand, 
its much more difficult for computers to decide if different simplified 
polygons are the same. Particularlly if different jurisdictions want to 
very tightly define warning areas.

Mobile phone networks measure over-the-air capacity in bits.  The more 
bits used for more detailed polygons, the fewer bits for other information 
in the warning message. Even with polygon compression, CMU was estimating 
45 bytes for a polygon with 20 points based on National Weather Service 
polygons. On the other hand, local storage is measured in terms of 
megabytes or gigabytes. This is an age-old trade-off since the beginning 
of computer and transmission networks.

Civil warnings for jurisdictions such as the "entire United States" or a 
state or county may be better described using a different system.  A 
polygon may work better for a simple geometric area or some natural 
hazards which move in simple ways. Other systems may be better for human 
hazards such as AMBER alerts and civil jurisdictions.

> It seems like you're trying to use a single instance to prove some 
> general principle, but I'm really not following your logic.  Anyway, if 
> you have a constructive suggestion I'm sure we all love to hear it.

Its hardly a single instance. The National Weather Service has several 
studies (and continues to fund academics for more studies) involving the 
trade-offs identifying warning areas using different systems. The studies 
regularly find that no single system is best for all types of hazards or 
warning scenarios.

I'm glad if you are open to suggestions and improvements rather than 
always defending CAP as the answer.  The OASIS Common Alerting Protocol is 
a transport standard. Giving someone a copy of the HTML protocol standard 
won't make them web page experts. Almost no one creates HTML by hand, and 
likewise I doubt many emergency managers create CAP/XML messages by hand.

While the simpliest devices might use geographic polygons in isolation, 
humans generally need some kind of map to make sense of a geographic 
polygon. High resolution maps consume gigabytes.  Low resolution maps 
consume megabytes. Whether high or low resolution, maps on smart phones 
and other consumer devices include metadata such as national and 
state borders.

What's the tradeoff between transmitting complex jurisdictional polygons 
with each warning message versus simple maps and metadata stored locally?

If warning messages have multiple location descriptions, i.e. geocodes, 
polygon, textual description; how should consumer alerting devices 
prioritize and de-conflict the different locations?  Use the broadest 
warning area, the most specific warning area, the average warning area? 
This is the biggest problem with NWS warning messages transmitted as 
CAP/XML messages.

The lack of written requirements and a written concept of operations makes 
things more confusing.  Just refering to transmission protocols like 
S.A.M.E, EAS and CAP/XML doesn't answer the questions.