[Requests] Feedback for [17-083, OGC Tile Matrix Set Standard Candidate]

Pyles, John John.Pyles at gtri.gatech.edu
Wed Jul 25 19:29:48 EDT 2018


PART A

1. Evaluators:

Vincent Sollicito
Principal Research Scientist
Georgia Tech Research Institute
404-407-6971
Vincent.Sollicito at gtri.gatech.edu

John H. Pyles, Sr.
Principal Research Scientist
Chief, Geospatial Technologies Branch
Georgia Tech Research Institute
404-786-8212 (mobile)
John.Pyles at gatech.edu<mailto:John.Pyles at gatech.edu>


2. Submission: [17-083, OGC Tile Matrix Set Standard Candidate]


PART B

Upon review of “OGC Tile Matrix Set Standard Candidate”, we found the following errors (most of which come from the WMTS standard), and have the following suggestions.

Likely Errors:

C.3 GoogleCRS84Quad – If you want a constant tile size of 256x256, then Level 0 is a special case with the whole world fitting in a single tile with the first 64 and last 64 lines being blank. That is, 256x128 pixels of data covering the world. Level 1 should be 2x1, Level 2 should be 4x2, and so on. This is contrary to, “The next level represents the whole world in 2x2 tiles of 256x256 pixels and so on in powers of 2”, which would imply transparent rows or tiles at every level. https://knowledge.safe.com/articles/180/which-web-map-tiling-scheme-should-i-use.html would be the preferred approach.

C.4 GoogleMapsCompatible and C.5 WorldMercatorWGS84 both indicate that they cover the whole world; however, the poles are at infinity in the Mercator projection. That fact, combined with the desire to have data set that fits nicely into a quad-tree topped off by a single, 256x256 pixel, Level 0 tile that covers the entire data set made the limits what they are (Web Mercator: +/-85.0511287798° latitude, World Mercator: +/-85.08405903° latitude).

D.3 World CRS84 Quad TileMatrixSet provides a non-conformant example. As stated in the OpenGIS® WMTS Implementation Standard,

“To be conformant to these well-known scale sets, a WMTS server SHALL allow responses from the largest scale denominator on the following tables and all intermediate scale denominators down to the most detailed scale resolution of that data; it is therefore not required to support the smallest scale denominators in order to be conformant to a well-known scale set.”

The example skips the top level listed in C.3 (scale denominator: 559082264.0287178, pixel size in degrees: 1.40625000000000). While D.3 NOTE2 acknowledges this omission, without the top level that covers the entire world in 256x128 pixels, this TileMatrixSet is incomplete and non-conformant. GTRICRS84Quad provides a conformant example.

Suggestions:

1.           Tile sizes should not be a part of the definition of a Well-known Scale Set. C.1 defines GlobalCRS84Scale and C.2 defines GlobalCRS84Pixel. In both cases the Tile Width and Tile Height are not specified. In the case of GlobalCRS84Scale, no integer tile size results in the exact scale denominators specified. In the case of GlobalCRS84Pixel, and exact solution exists using a Tile Width of 180 and Tile Height of 90. However, there is no reason why a server couldn’t use larger tile sizes at Level 1, 2, etc. There is nothing in the standard that says you have to use a single tile size at every level, nor is it required that all tiles be 256x256, even if it means leaving half the pixels empty, as was done with GoogleCRS84Quad Level 0. Each TileMatrix in a TileMatrixSet can specify its own TileWidth and TileHeight, and still be compatible with the WellKnownScaleSet. Larger tile sizes means fewer HTTP requests and fewer files on disk. A TileMatrixSet that uses a 1024x1024 tile size can deliver in one response, the same data that requires 16 requests for data organized in 256x256 tiles.

2.           GoogleCRS84Quad based Tile Matrix Sets should be the preferred / recommended choice for world wide data sets. Unlike WorldMercatorWGS84 or GoogleMapsCompatible / WebMercator, GoogleCRS84Quad actually covers the entire world. While both Equirectangular and Mercator projections have East-West distortion, the further you get from the equator, Mercator also stretches features in the North-South direction. For a worldwide data set, at every WMTS level, a WorldMercatorWGS84 or GoogleMapsCompatible Tile Matrix will have twice as many pixels relative to the same level in GoogleCRS84Quad. GoogleCRS84Quad is easier to work with, whether you are converting data in another datum or projection into a Tile Matrix Set, or you are consuming a Tile Matrix Set and presenting it in a different projection.

3.           Azimuthal Equidistant (Polar Aspect) should be the preferred / recommended choice for polar data sets (see Polar Azimuthal Equidistant tab in WMTS Well-known Scale Sets.xlsx). CADRG, CIB and ECRG use this projection for the Polar Zones. Both the North Polar Aspect and South Polar Aspect can be worldwide. The WMTS levels and scales line up nicely with those of GoogleCRS84Quad, WorldMercatorWGS84 and GoogleMapsCompatible (WebMercator).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengeospatial.org/pipermail/requests/attachments/20180725/b0832f9d/attachment.html>


More information about the Requests mailing list