[EAS] USVI and PR: polygons versus political boundaries

Wed Sep 6 23:02:05 CDT 2017

Sean,

>Trying to draw polygons between the U.S. Virgin Islands and the British
>Virgin Islands can create some very oddball warning boundaries.
>Sometimes its clearer to just reference the political subdivisions.

Depends on what you mean by "oddball," doesn't it?  GIS folk do that sort of thing all the time, down to precisions of meters.  Yes, it's easier for a human brain, which is optimized for language rather than geometry, but that's not the target audience.  The goal is to enable computerized devices to determine whether a particular lat/lon location (often from GPS) is inside or outside the target area.

Not everything that's unfamiliar is bad, Sean.

A FIPS code for a state or a county is really nothing but a shorthand for a particular polygon... and in fact it's possible to create a lookup table to get from FIPS to a corresponding polygon.  But since the computation is based on a polygon anyway, why not use one that actually represents the hazard rather than just some political construction from a hundred years ago?

Here, for example is a simplified poly around the USVI (in GeoJSON format).  This is from the lookup database behind http://warn.pbs.org, which is a realtime map of WEA activity nationwide:

{ "type": "Feature", "properties": { "STATEFP": "78" }, "geometry": { "type": "Polygon", "coordinates": [ [ [ -65.046395965276687, 18.382517500826904 ], [ -64.980067016510077, 18.396303733127176 ], [ -64.953047678654158, 18.403396139364911 ], [ -64.920893984082909, 18.407710871978498 ], [ -64.906382173394448, 18.407334350950485 ], [ -64.816496150443825, 18.387862521207953 ], [ -64.815869483725393, 18.390755345343131 ], [ -64.744532140634831, 18.371908187738786 ], [ -64.736455117923043, 18.370656571440698 ], [ -64.709373946789654, 18.364111790399633 ], [ -64.674931718294104, 18.350977555176058 ], [ -64.668860069643102, 18.346783788355676 ], [ -64.666122729918641, 18.344301646424725 ], [ -64.66551786811111, 18.341789223148751 ], [ -64.666514581456852, 18.334084408815816 ], [ -64.664887464126807, 18.065491732026498 ], [ -64.591829246071498, 17.8070209013507 ], [ -64.570409316386886, 17.757761527316497 ], [ -64.59354955894014, 17.741373084846131 ], [ -64.602153300776905, 17.737260460173477 ], [ -64.622033176376931, 17.729317765823769 ], [ -64.677048054000096, 17.712000285514023 ], [ -64.697969733856169, 17.705160631247566 ], [ -64.708544653189904, 17.701574214192036 ], [ -64.749616271821779, 17.690761961551885 ], [ -64.762745612220556, 17.686947128091248 ], [ -64.798278621751138, 17.683208005357464 ], [ -64.834977471070857, 17.682203600034221 ], [ -64.883003071380742, 17.679732479216145 ], [ -64.89805835918024, 17.678594309705041 ], [ -64.899072399040634, 17.680220830126721 ], [ -64.900588739458996, 17.684179820618279 ], [ -64.900165346780213, 17.690697334267721 ], [ -64.926973363491598, 17.99424239750131 ], [ -65.043954991537149, 18.275620388597435 ], [ -65.07993206000701, 18.334702139156271 ], [ -65.078779356587845, 18.340053529328294 ], [ -65.076957133431293, 18.345034936226909 ], [ -65.076592161051806, 18.345785292701311 ], [ -65.063412131033374, 18.374440401354352 ], [ -65.062199352683777, 18.37616260947415 ], [ -65.061336907316871, 18.376791888783526 ], [ -65.055394701902102, 18.37972304265206 ], [ -65.046395965276687, 18.382517500826904 ] ] ] } }

The polygon is simplified by "buffering" (expanding the boundary outward) and then "unbuffering" it to reduce the number of distinct points.  Polities with water boundaries... oceans, lakes, rivers, even the occasional creek... tend to have extremely high number of points in their precise boundaries.  By convention CAP limits polys to 100 points (vertices) as hazard boundaries are very rarely that complex and political ones can be simplified.

>I have no idea what WEA cell towers would be included or excluded from
>this warning polygon versus just referencing Culebra (72049) and Vieques
>(72147).

 Why should you?  As an alert originator your responsibility is in the hazard and the message, not the delivery technology.

The cellular companies consider that level of network detail highly proprietary... and in any event it changes daily in the course of network maintenance and expansion.  Nonetheless, the carriers do know, and routinely use polygons (representing exactly what, they neither know nor care) to select sites to activate. 

Note that not all wireless carriers do that exactly the same way.  Some look up which cell sites are inside the polygon, others look up all the cell sites whose coverage boundaries intersect the polygon, and so on.  That's one of the things that makes evaluating WEA performance tricky... the same message will go to somewhat different areas on different carriers.  (Yeah, that sucks, but you try negotiating with the big wireless carriers!  We're working on it.)

That's why there's discussion of using the GPS and network-derived location awareness of individual handsets to do the calculations.  If you'd like to ask the handset manufacturers to make room for a complete FIPS database on every phone, just to preserve the old ways, by all means go for it!

>Sometimes polygons work best. Sometimes civil identifies work best. Being
>able to use more than one geo-location system allows different choices for
>different situations.

Um... not really.  Some systems (computers) handle polygons best.  Some systems (humans) handle names best.  If we ever build a system where some human has to pick out the transmitters to use by hand, then maybe that'll be a case where string labels are best.

Think that's going to happen anytime soon?

Art