Spatial join distance between polygons and points

[annfeng1]

I can't seem to find an answer from the documentation. When joining points to polygons to choose points closest to each polygon, the spatial join tool gives this message that it's choosing points closest to the boundary of polygons. Should i take this to mean that the tool is calculating distance between points to the closest edge of the polygon, not the polygon centroid? If so, it accomplishes the same task as the near tool? I thought spatial join is an overlay type of function, not proximity-based, and hence distance computing uses centroid of each polygon. Any clarification is very much appreciated!

ann

[DanPatterson_Retired · Accepted solution]

It is not to the centroid.  The graphical examples Select By Location: graphic examples—Help | ArcGIS for Desktop indicate the types of relationships when the various geometries are involved

[annfeng1]

Thanks Dan for your speedy reply. Suppose a feature class isn't projected, the distance is then given in decimal degrees? Based on the impossible results I am seeing, I guess on-the-fly projection doesn't accomodate spatial join distance calculations and therefore we have to project the target feature first? Thanks again for any pointers!

[DanPatterson_Retired]

I always recommend that the geometries be projected since you want to ensure that the best projection is used to ensure proper results in projected units.  The section in the spatial join help file explains this a bit more How proximity tools calculate distance—Help | ArcGIS for Desktop

There is nothing wrong with having your data in unprojected and projected coordinates.  I just keep them in separate data frames.  If you are working with a projected coordinate system, like UTM, then your distance calculations will be more than accurate for most purposes.  Regardless, the closest feature is chosen, but my recommendation is to used projected data to get planar units like meters, kilometers etc.  It takes the guess work out of 1.125 being km or is it minutes of a degree ? ? etc

[annfeng1]

Thanks again Dan for the helpful tip! Incidentally, can I pick your brains on what UTM zones are the usual choice for say population centers in western Canada e.g., Edmonton,Calgary. They tend to span multiple zones. There appeared to be some custom projected coordinate systems like the 3 degree Transverse Mercator(?) Sorry I am veering off topic a bit. Do you have suggestions on the projected coordinate systems for the above mentioned metropolitan areas? Thanks!

[DanPatterson_Retired]

We use the Modified Transverse Mercator here (Ottawa) as well (3 degree MTM).  It is totally acceptable.  If the place you are located spans or is split by a UTM zone, then use the designated projection for your area so that your data will fit in.  If you want to cover a larger east-west direction, then there are Lambert Conformal Conic projections that are used.  For example, Ontario is split by several UTM zones, so a LCC for Ontario is used instead.  Toronto and Ottawa fit within one UTM zone so you can use UTM 18 to cover that east west extent.  In Ottawa, you can use either a UTM, or MTM projection, but Ottawa's MTM wouldn't cover Toronto area.  It is all a matter of East-West extent in the case of these 3 projections... choose what best fits all the data you are using

[annfeng1]

Thannks Dan! You have been most helpful.

[DarrenWiens2]

Search for "Alberta 10tm" for the standard Alberta-wide traverse Mercator projection. It doesn't come out of the box, so you'll have to make it custom.

[annfeng1]

Thanks for your tip!

[annfeng1]

Thanks Dan for your speedy reply. Suppose a feature class isn't projected, the distance is then given in decimal degrees? Based on the impossible results I am seeing, I guess on-the-fly projection doesn't accomodate spatial join distance calculations and therefore we have to project the target feature first? Thanks again for any pointers!