[Qgis-user] Question on Project Properties - Ellipsoid

[Randal Hale]

On 7/13/20 7:22 PM, Nyall Dawson wrote:
> On Tue, 14 Jul 2020 at 05:59, Randal Hale
> <rjhale at northrivergeographic.com> wrote:
>> I hope everyone is staying safe out there.
>>
>> So I've had some questions regarding setting up a QGIS Project and now
>> I've convinced myself that I'm doing this wrong. When I get a new
>> project in and I start setting up QGIS generally my project will be in a
>> Stateplane projection - in this case EPSG:2274 (TN Stateplane NAD83).
>> When I set up the project I change the ellipsoid to be
>> "none/planimetric" (In the project properties). I don't set up the
>> ellipsoid to be GRS 1980 and haven't for some time.
> Can I ask why you do this?
That's a great question - somehow I became convinced this was right. 
Mostly it resulted from a job I was doing for a forester - the area 
calculations from ellipsoid were coming out larger than it should have 
been. Planimetric put me close to what was "right". I started doubting 
this was right a few weeks ago wrestling with the "ballpark transform" 
error message.
>> Now I'm getting a
>> ballpark transformation message if I move data from 2274 to 4326 (for
>> instance I'm moving points to a GPX Format).....BUT - if I set the
>> Ellipsoid to be GRS 1980 (which it is for that projection) - no warning.
> Can you open a ticket on this and attach your data? It's quite
> possible that the warning is correct, but it's also possible it's a
> spurious warning and I can hide it in this circumstance.
I can!
>
>> Which now has me thinking MAYBE I should set the ellipsoid..... EXCEPT
>> my area measurements always come out a bit larger than expected.
> Which **IS** expected! :D An ellipsoidal area will almost always
> result in a larger area vs a cartesian area calculation, because it's
> calculating the area over a curved surface rather than a flat surface.
> (same applies to length measurements).
>
> Generally my advice is to use ellipsoidal area/length calculations,
> unless the specific constraints and specifications of a particular
> task or discipline dictate otherwise. If in doubt, I'd:
>
> 1. compare the two area calculation results. The identify tool
> "derived attributes" section handily shows BOTH cartesian and
> ellipsoidal areas/lengths, if you've setup your project to use an
> ellipsoid. If the results are relatively similar (e.g. say within 0.1%
> of each other), then ask yourself whether it makes any difference
> practically which of them you use. Chances are the accuracy of your
> input datasets will be a bigger factor here then the area calculation
> method used, and you may be chasing false accuracy by reporting areas
> to that many significant figures anyway! Or, in other cases, the
> actual purpose of the calculations you're making aren't even required
> to a degree of accuracy to warrant being affected by the choice!
> (e.g. if you're calculating species density, then it's an inherent
> estimate only and the practical outcome of choosing between
> cartesian/ellipsoidal calculations is nill).
I'm really starting to wonder if this started back at 3 (in my head) and 
I've continued down the wrong path since then. Maybe it was a bug or 
something in the beginning and I never adjusted based on everything 
getting fixed. Maybe. I really don't have a good answer.
>
> 2. if you've checked and the input datasets ARE sufficiently accurate
> that the choice matters, then push the choice back to your customer.
> Ask them which method is suitable for the results they're after. Get
> the answer in writing and your method is justified!
>
> 3. keep in mind that even ellipsoidal measurements are an estimation.
> The earth isn't an ellipsoid! Local variations in elevation are likely
> to have a much more significant impact on large scale calculations vs
> the cartesian/ellipsoidal choice.
>
> Hope that helps!
Yes - it helps. I'm getting a lot of questions from people getting 
"ballpark transform" messages which led to the last little bit of 
testing, head scratching, and "Hey I can be completely wrong on what I'm 
doing".
> Nyall
>
>
>
>
>
>
>
>> So I go
>> back to "none/planimetric".....BUT - is that causing issue with
>> reprojecting data........
>>
>> Any thoughts? Thanks for any support or pointing me in a direction.
>>
>> Randy
>>
>> --
>> Randal Hale
>> rjhale at northrivergeographic.com
>> (p) 423.653.3611
>> http://www.northrivergeographic.com
>>
>> _______________________________________________
>> Qgis-user mailing list
>> Qgis-user at lists.osgeo.org
>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-user
>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user

-- 
-----------
Randal Hale
rjhale at northrivergeographic.com
https://www.northrivergeographic.com
(423)653-3611