Georeferencing a TIF using multiple points

Question

Hi,

I need to do an analysis of an area somewhere in Madagascar. I do not have extensive data to start the analysis, but I have a map in a pdf format. The map shows a grid that is divided well enough that I can pinpoint every degree of latitude and longitude. I will need to add elements on the map to do my analysis and require precise location. These elements are defined in WGS84. 
So here is the problem.

I created a shapefile of points for every lat and long of an area on the map, and imported it in ArcMap 9.2
I saved the pdf file in a TIF format and imported it in ArcMap 9.2 for georeferencing.
When I start to georeference the map with the points, the result is not bad but far from being precise. I tried to georeference each latlong with the points but it is a disaster.

Perhaps the grid on the map is not perfectly divided or I can't stretch the map properly ?
(Perhaps it is the user?)
The map does not tell the projection or coordinate system which doesn't help.

Something I find funny, is that if I use the ruler to calculate the distance between each degree of latitude, I get 59.75 NM where it should be 60 NM. These are the points from the shapefile in WGS84. Does not make sense to me.

By the way, I never installed the service packs for 9.2, and will get ArcGIS 10 soon. Don't know if this will make a difference.

Thanks for any help I can get.

JacquesBeaudry · Answer

Thank you so much for your input.

I just ordered georeferenced maps for my project and perhaps I did something wrong but time will tell. I've seen Madagascar maps using the Laborde projection, but I did not see this option in version 9.2, so I ordered the Lambert Conformal conic projection. My world of expertise is aeronautical charting and we often use this projection. I hope I will be able to match my database with these maps.

I will get version 10 within the next month and hope it will help. Meanwhile I think I need to understand projection a lot more as I don't know how ArcMAP does the job. As an example, how come I can use a georeferenced map in NAD83, create a shapefile using the same system, and be able to draw a line with the precise length (ex: 3 NM) and some other times it does not work. I feel like a beginner when I see things like that.

Thanks again, and please feel free to add any comments

Ciao!

MelitaKennedy · Answer

Hi,

I need to do an analysis of an area somewhere in Madagascar. I do not have extensive data to start the analysis, but I have a map in a pdf format. The map shows a grid that is divided well enough that I can pinpoint every degree of latitude and longitude. I will need to add elements on the map to do my analysis and require precise location. These elements are defined in WGS84. 
So here is the problem.

I created a shapefile of points for every lat and long of an area on the map, and imported it in ArcMap 9.2
I saved the pdf file in a TIF format and imported it in ArcMap 9.2 for georeferencing.
When I start to georeference the map with the points, the result is not bad but far from being precise. I tried to georeference each latlong with the points but it is a disaster.

Perhaps the grid on the map is not perfectly divided or I can't stretch the map properly ?
(Perhaps it is the user?)
The map does not tell the projection or coordinate system which doesn't help.

Something I find funny, is that if I use the ruler to calculate the distance between each degree of latitude, I get 59.75 NM where it should be 60 NM. These are the points from the shapefile in WGS84. Does not make sense to me.

By the way, I never installed the service packs for 9.2, and will get ArcGIS 10 soon. Don't know if this will make a difference.

Thanks for any help I can get.

The measure tool could be measuring the geodesic distance.

I've noticed the same issue--there's a certain point where adding more control points doesn't help the least squares fit (in the georeferencing).

The map could be using the Laborde projection which is the national standard. We didn't support until version 10. You might trying converting a few points to Laborde and georeferencing to them.

Melita

JohnSobetzer · Answer

Your best hope for good advice is for Melita Kennedy to answer.  But here's some thoughts you might consider in the interim.  It would really help if you knew the coordinate system the pdf map was in.  If that isn't something you can get and you have some time to play, you might try some quickie georeferencing with your data frame in some different projections, perhaps ones commonly used in your area.  See if one seems to fit particularly well.  If there is then start afresh and use it.

If your map has some obvious features spread around it, you might consider using a photo or good vector layers to attempt a georeferencing.  Road intersections, lake outlets, structures, etc. are all good candidates for control points.

If when you get done you can't simple stretch and rotate your map, look at the other transformations options and see if one is better for your work than others.