Subscribe to this thread
Home - General / All posts - Optimal projection for distance and area calculations
boomerbubba15 post(s)
#24-Jul-14 14:45

I am trying to decide on a default projection to use for storing map data for Texas statewide. I have two choices: EPSG 3083 based on Albers Equal Area, or EPSG 3084 based on Lambert Conformal.

Sometimes area would be significant for my application. So I am leaning toward the Albers-based projection.

But much of the functionality I need to support would be distance-oriented. For example, nearest-neighbor and Voronoi analysis. Is the Albers projection an acceptable choice for that?

volker

1,086 post(s)
#24-Jul-14 15:12

Have a look at Manifold Help at:

Guide to Selecting Map Projections

........

Addendum: Suggestions from the Manifold Team

Mapping experts will choose from a wide array of projections to meet specific objectives as discussed above. For general use, most people will usually develop a favorite projection for a particular situation and then use that projection over and over. For most situations either Orthographic or Lambert Conformal Conic are good choices.

The Manifold team recommends the following:

Region to be Mapped

Suggested Projections

Whole Earth

Robinson (pseudocylindrical) or Miller Cylindrical . Robinson seems to be fashionable for thematic maps. Any of the pseudocylindrical projections will be fine if you like their appearance better.

Hemispheres

Orthographic (azimuthal) for a "view from space" look, and Lambert Azimuthal Equal Area for thematic maps where the relative size of countries near the edge of the projection is to be preserved.

Continents

Use Lambert Conformal Conic for North America and Eurasia. Use Lambert Azimuthal Equal Area or Orthographic for South America and Africa. Use Orthographic for Australia, and Antarctica.

E-W Countries or Regions

Use Lambert Conformal Conic for US, Canada, Russia, and China. Use either Lambert Conformal Conic or Orthographic for Europe. Use Orthographic or Lambert Azimuthal Equal Area otherwise.

Polar Regions

Orthographic or Lambert Azimuthal Equal Area.

Oceans

Orthographic or Lambert Azimuthal Equal Area.

Smaller Countries or Regions

Orthographic.

N-S Countries, Oblique Regions

Long, thin countries aligned North-South such as Chile are one of the few times we would use Transverse Mercator . Oblique regions like the Alaska panhandle are mercifully rare: Use the Oblique Mercator in such cases.

Other than personal taste in visual appearance there are only two reasons not to use one of the above projections:

  • A specific projection is required to match other data available or to exchange data with other users.

  • The technical characteristics of a particular projection such as preservation of scale or relative area are required for creating a paper map or other presentation.

Manifold provides a very wide array of projections in addition to the standard projections mentioned above


http://www.thegisservicesector.de

boomerbubba15 post(s)
#24-Jul-14 15:27

Thanks. I had already read that help section but it did not resolve the question for me.

BTW, I am not so worried about changing projections on the fly, if necessary, within Manifold. But I also am looking to store this spatial data on SQL Server 2012. It would be useful to perform some operations on the server natively without sucking it into my Manifold desktop, so the "default" projection I am looking for really refers to how I store the data on the server.

So another option for me is just to use the SQL Server geography datatype with the SRID of 4326. I would do that, or project my data in Manifold first using either the Albers Equal Area or Lambert Conformal projection, then upload it to SQL server as a flattened geometry datatype.

tjhb
10,094 post(s)
#25-Jul-14 08:46

Nice question. Are you definitely limited to a choice between those two projections (or latitude/longitude)?

Another projection worth considering is a Lambert Azimuthal Equal Area centred near a weighted centroid of the state of Texas (scale will be compressed somewhat towards the edges of the state, but symmetrically about the centre). And a standard UTM (or custom Transverse Mercator) would also be worth a thought--a good compromise between area and distance.

My intuition is that conformality is also a significant consideration for Voronoi segmentation and (therefore) nearest neighbour analysis, especially if the points are widely spaced.

It would be an interesting exercise to assess all these projections (and maybe others) systematically for Texas.

boomerbubba15 post(s)
#25-Jul-14 14:23

I work for a state agency. While there is not an ironclad requirement, the two projections I mentioned in the OP are strongly favored in standards adopted by the legislature for statewide mapping. See this handy explanation from Texas A&M curriculum notes:

http://geography.tamu.edu/class/aklein/geog390/lecture/texas_map_projections.html

I think the intent of that really applies to projections used for publishing data. Of course I am free to store it and manipulate it internally however I like.

For my analytical purposes, perhaps it would be ideal to use the SQL Server geography datatype for purposes of area calculation, nearest-neighbor queries and Voronoi analysis. (The latter would be challenging because SQL Server has no built-in functionality for that.) Perhaps I am worrying about optimization too much. Even though Texas spans a wide area, most of the data points I think I would need to use for nearest-neighbor matches and generating Voronoi areas would probably not be widely spaced.

BTW, I also will have some point-in-polygon queries to execute. But as far as I know the choice of projection is not critical for that purpose.

Manifold User Community Use Agreement Copyright (C) 2007-2021 Manifold Software Limited. All rights reserved.