[Proj] MGRS notation in extended UTM zones?

[Charles Karney]

On 06/21/11 07:35, Mikael Rittri wrote:
> Thanks, Charles.
>
> I suppose this means that there are no official guidelines,
> unless one would accept GeoTrans as an authority. But I don't
> have much confidence in GeoTrans (although I haven't
> investigated it in detail, as you have).
>
>> In the east-west direction this means that UTM eastings
>> in [100km,900km) are legal.
>
>> The rule for the latitude band letter is that if a latitude
>> band boundary intersects a 100 km MGRS block then either
>> letter is valid for any coordinate in that block.
>
> I agree that the second rule is necessary (more or less) and harmless.
> But I am less sure about the first rule.
>
>> I can expand on my reasons for these choices if anyone's interested.
>
> Yes, I would be interested in more details.

Motivations for the rules for allowed MGRS conversions:

(1) Repeated conversions should be stable.  Given that UTM/UPS -> MGRS
conversion is by truncation, then MGRS -> UTM/UPS should return the
center of the square (at the appropriate precision).

(2) Truncating a legal MGRS coordinate should result in a legal MGRS
coordinate; similarly adding precision to an legal MGRS coordinate
should result in a legal MGRS coordinate.

(3) MGRS lettering boundaries should be adhered to since these supply
redundancy checks.  Thus zone 02 can only use column letters J-R.  It
shouldn't be extended further east by using letter S, because zone 03
uses column letters S-Z and doing so would lose a check.  Similarly the
use of latitude band letters should be as strict as possible consistent
with (2).  (These rule allow various bits of the MGRS code, e.g., the
UTM zone number, to be dropped with the nearest point of possible
confusion being reasonably far away.)

A consequence of rule (2) is that the extent of a numeric zone should
follow the boundaries of the 100km MGRS blocks.

The maximum extent of a UTM zone [100km,900km] in the E-W direction (due
to rule 3), and GeographicLib imposes a limit in the N-S direction in
order to guarantee a 0.5 deg overlap with UPS.  This defines a
rectangular regions in projected space for each UTM zone (and similarly
for the S and N UPS zones).

The only remaining question is to make the legal regions smaller than
these in order to make the overlap region no more than about 0.5 deg.
The boundary would then zig-zag to follow the 100km block boundaries.  I
elected not to do this in GeographicLib, because

(a) this would be tricky to get right (especially with the exceptions
for Norway and Svalbard)

(b) there's nowhere I can since in the US DoD documentation that says
that the overlap should be restricted to 0.5 deg.  GeoTrans once (and
may still) enforced this overlap restriction but that's for a bad
reason: that the GeoTrans UTM algorithm is inaccurate much beyond 3 deg
from the central meridian

(c) the sort of coordinate transformations I need to do, e.g., integrate
Lidar collects in different UTM zones, a stricter overlap region would
inevitably "get in my way".  Being generous allows, for example, a user
to use "doubled" UTM zones in Greenland and the north of Canada similar
to the double-wide Svalbard zones.

(d) if restrictions are needed then they can be enforced by a separate
layer outside GeographicLib.