Help identifying coordinate system

Barrydriscoll · ‎08-07-2013

Hi all,

I have what I hope is a simple problem from an ArcMap novice.

I have two sets of shapefiles for a country. One set is shapefiles for roads and lakes within the country. The other is one shapefile for the country boundary. The set with country features (shapefiles for roads, lakes, etc) all have defined coordinate systems and overlay just fine on the fly. The country boundary does not have a defined coordinate system and that's where I'm stuck.

The extent values for the country boundary are shown in this picture. From what I've read, I'm guessing the original coordinate system is projected rather than geographic (based on the size of the values).

[ATTACH=CONFIG]26559[/ATTACH]

I read online that one can use some trial and error and keep defining projections until you hit on the original coordinate system but I had no luck with this.

If I'm right in thinking that my steps are (1) define the projection of the country boundary shapefile, and then (2) project it to overlay with the road & lake shapefiles, is there something I'm missing in order to do step 1?

In case it helps, here are the details of the lake and road shapefiles. All the shapefiles (including my problematic one) come from the same office.

---

Extents
Top: 756142.379950 m
Right: 1142656.619454 m
Left: 690629.179290 m
Bottom: 74790.443758 m

Projected Coordinate System: Ghana TM
Projection: Transverse_Mercator
False_Easting: 900000.00000000
False_Northing: 50000.00000000
Central_Meridian: -1.00000000
Scale_Factor: 1.00000000
Latitude_Of_Origin: 4.66666667
Linear Unit: Meter

Geographic Coordinate System: GCS_Clarke_1866
Datum: D_Clarke_1866
Prime Meridian: Greenwich
Angular Unit: Degree

----

Top: 769664.186871 dd
Right: 1142674.650690 dd
Bottom: 57841.861131 dd
Left: 650523.913583 dd

Geographic Coordinate System: GCS_WGS_1984
Datum: D_WGS_1984
Prime Meridian: Greenwich
Angular Unit: Degree

---

Any help greatly appreciated!

MelitaKennedy · ‎08-07-2013

The 2nd extent looks very similar to the first. What happens if you redefine (not reproject) the 2nd set to use the same coordinate system?

If that doesn't work. Can you overlay the first set (that's using Ghana TM) with a 3rd set of data like general country boundaries? The Ghana TM definition that you have doesn't match other definitions that I've seen for Ghana data.

Melita

Barrydriscoll · ‎08-07-2013

The 2nd extent looks very similar to the first. What happens if you redefine (not reproject) the 2nd set to use the same coordinate system?

If that doesn't work. Can you overlay the first set (that's using Ghana TM) with a 3rd set of data like general country boundaries? The Ghana TM definition that you have doesn't match other definitions that I've seen for Ghana data.

Melita

Thanks Melita.

No luck so far.

Your first question: Here's what happens when I redefine the country boundary (the one with no coordinate system) using the coordinate system from the road shapefile:

[ATTACH=CONFIG]26560[/ATTACH]

Your second question: I didn't have a problem overlaying the set using Ghana TM with a map of West Africa I found online. This picture shows a shapefile with powerlines using 'Ghana TM'. I started with a West Africa map (GCS_WGS_1984) and had no problem adding the one using Ghana TM.

[ATTACH=CONFIG]26561[/ATTACH]

Where do you think this leaves me with my undefined boundary shapefile?

Thanks for your help.

MelitaKennedy · ‎08-07-2013

One thing I noticed when I looked up Ghana coordinate systems, and that led me to ask whether the first set could be overlaid over other data, is that there's a version that uses a "Gold Coast foot". It certainly looks like the layer is 1/3 the size of the other data! Leave it undefined. Add and the other data to ArcMap. Try setting the data frame's coordinate system to Accra Ghana Grid and see if they line up. You should get a gcs/datum warning message.

Melita

curtvprice · ‎08-07-2013

Warning, book plug for a friend... and, honestly, this is just a neat book that helps solve just this problem. If you find yourself puzzling coordinate systems often in your office, you may find it very helpful:

Lining Up Data in ArcGIS: A Guide to Map Projections (Margaret Maher, Esri Press, 2010)

I'm hoping they let her do another edition that incorporates huge improvements (hey, thanks Melita!) to coordinate system handling in 10.1.

TOMKONG · ‎08-07-2013

---

Extents
Top: 756142.379950 m
Right: 1142656.619454 m
Left: 690629.179290 m
Bottom: 74790.443758 m

Projected Coordinate System: Ghana TM
Projection: Transverse_Mercator
False_Easting: 900000.00000000
False_Northing: 50000.00000000
Central_Meridian: -1.00000000
Scale_Factor: 1.00000000
Latitude_Of_Origin: 4.66666667
Linear Unit: Meter

Geographic Coordinate System: GCS_Clarke_1866
Datum: D_Clarke_1866
Prime Meridian: Greenwich
Angular Unit: Degree

----

Top: 769664.186871 dd
Right: 1142674.650690 dd
Bottom: 57841.861131 dd
Left: 650523.913583 dd

Geographic Coordinate System: GCS_WGS_1984
Datum: D_WGS_1984
Prime Meridian: Greenwich
Angular Unit: Degree

---

Any help greatly appreciated!

I'm wondering about above info you provided:
1.for the first feature class, the projection is TM (it's scale factor is 1.0000000--should be small area UT, 1degree TM?), the GCS is labelled as GCS_Clarke_1866. But the Clarke_1866 only is the referenced ellipsoid model and it is not the Datum (ig. The datum NAD27 -- its referenced ellipsoid is Clarke_1866, but its datum is not Clarke_1866),what is its datum?
I thought the first one missing datum--some one put the wrong datum info.
2.the second feature class doesn't have the labelled projection.

So both layers may have issues.

Barrydriscoll · ‎08-08-2013

One thing I noticed when I looked up Ghana coordinate systems, and that led me to ask whether the first set could be overlaid over other data, is that there's a version that uses a "Gold Coast foot". It certainly looks like the layer is 1/3 the size of the other data! Leave it undefined. Add and the other data to ArcMap. Try setting the data frame's coordinate system to Accra Ghana Grid and see if they line up. You should get a gcs/datum warning message.

Melita

Thanks Melita,

That seems to have worked!

Here's the result of overlaying several shapefiles. I began with the undefined boundary shapefile (Districts), then set the data frame to Accra Ghana Grid, then added shapefiles for Africa, and then shapefiles specific to Ghana (regional, lake, airport, railroad).
[ATTACH=CONFIG]26576[/ATTACH]

As is, the problematic shapefile (the District boundary) remains undefined, the airport, railroad and lake shapefiles are Ghana TM and GCS Clarke 1866 with extents in meters, and the regional and Africa shapefiles are GCS WGS 1984 in dd.

Thank you very much. I never would have guessed I needed to set the data frame to Accra Ghana Grid!

@curtvprice - Thanks for the recommendation.

@TKONG - do you think the problems you identified could cause me difficulty later on?

TOMKONG · ‎08-08-2013

@TKONG - do you think the problems you identified could cause me difficulty later on?

Depends, some softwares read the datum parameters but name, some softwares only read the datum name.
The ArcGIS reads the datum parameters (ellipsoid parameters), so it is not a big issue for ArcGIS if the datum name is wrong but its parameters are correct.

I mat one situation: its datum name is correct (NAD83), but the developer put the wrong ellipsoid model (parameters) for it (put Clarke 1866), so when this layer is added to the ArcMap, the ArcGIS read its datum as NAD27 but NAD83 even if its name is NAD83.
So the result is there are some distances between two datums (NAD27&NAD83) for same features in one ArcMap data frame, and people think the datum difference will cause location shift -- it is not right as the ArcGIS has "project on fly" function.

MelitaKennedy · ‎08-08-2013

What you should do now is define the districts data as Accra Ghana Grid using the layer's property page in ArcCatalog or the Define Projection tool.

Tom is correct that the spheroid-based GCS/datum of the first set of data could cause problems later.

If you continue to work at a relatively small scale, any offsets (correct or not) due to this incorrect definition won't be noticable.

My guess is that the first set of data came through ArcInfo workstation at some point as coverages or grids. ArcInfo Workstation uses a different 'projection engine'. If the datum isn't set, it defaults to Clarke 1866. When that definition is then converted to a shapefile or feature class, you get GCS_Clarke_1866. Because the rest of the definition doesn't match Ghana Grid, it's hard to say if the data is really based on Accra or Leigon or something else.

If you have some location where it's easy to overlay features to see if they line up, you might try modifying the coordinate system by changing Clarke 1866 to Accra (leave everything else the same) and see how they line up. If they're offset, try Leigon, set the data frame to WGS84, and set the geographic/datum transformations to convert from Accra and Leigon.

Melita

curtvprice · ‎08-08-2013

Usually when I see "Clarke_1866" specified in an old archived dataset it's a red flag that it was never set -- Clarke_1866 was often used in the GIS dark ages as a default when no datum was specified.

As you continue with your data rescue, you may find this article from PERS useful. (A plug for the Society - joining ASPRS is a good investment!)

The Republic of Ghana (PERS, June 2000)
PERS archive: Grids & Datums