Reprojection Issues

5243
12
Jump to solution
04-12-2015 09:08 AM
New Contributor II

Hi all

I'm hoping someone can help me out with some projection issues.  I have a 2006 National Land Cover Dataset (NLCD) for Louisiana originally in Albers Equal Area Projection.  The MRLC Server provided me two seperate tiffs to cover the data request for my study area (coastal Louisiana).

I mosaicked the two datasets into one, and reprojected both the mosaicked layer and my second dataset to the NAD 1983 South Louisiana State Plane Coordinate System.  I used the "Transformation" button under the Data Frame Properties tab to reproject, and I have also used the "Define Projection" tool in Arc Toolbox.  I even used the Raster Reproject tool, but the two layers still do not match up in Arc!  Please help!

Thank you!
~Ashley

Message was edited by: ashley tarver

Reply
0 Kudos
1 Solution

Accepted Solutions
Occasional Contributor III

Albers, UTM, and Louisiana State Plane South are all in NAD83 so there are no transformations to worry about.  Given that your data crosses two UTM zones, this system should probably be avoided.  Limiting the choices to  those that have been discussed, either Albers or Louisiana State Plane South should work fine for your needs.  I am not a coordinate system expert but both of these systems are widely used for spatial analysis.  Given that your data would be closer to the origin of the Louisiana State Plane South system, it should be susceptible to less distortion versus a continental system like Albers, which is one of the reasons local coordinates systems are always being pushed.  BUT I am not an expert and cannot confirm that this logic always holds true or applies at all in this case.  Going as local as is reasonable is just a rule of thumb I try to follow (since I am no expert and need "rules of thumb").  Seems to me they are both based on very similar projections so I don't think you can go wrong here.  You could always do some testing by projecting the cypress dataset into each of the prospective systems and running area calculations to see how much the results differ.  If you are seeing large differences in area based on these tests then coordinate system selection may warrant further consideration.

Lambert Conformal Conic (used for state plane):  ArcGIS Help (10.2, 10.2.1, and 10.2.2)

Albers Equal Area Conic:  ArcGIS Help (10.2, 10.2.1, and 10.2.2)

View solution in original post

12 Replies
Frequent Contributor

Hi Ashley, define projection is for defining the original projection. You don't use it to change a projection, nor should you do that in data frame properties, that just sets it for your data frame, not the actual datasets. You need to use the Project Raster tool to change a projection, and the Project tool if it is a feature dataset.

What was the original projection of your second dataset? It sounds like you might have defined that wrong to begin with. Did they line up before you attempted the reprojection? If not then then most likely it was defined incorrectly. Once you have the original projections defined correctly, using define projection, they should line up, and then you can project them into a new projection.

Frequent Contributor

When should I use the Define Projection tool and the Project tool?

I wasn't sure if I explained that as well as the above document does.

Reply
0 Kudos
New Contributor II

Hi Sephe-

Thanks for your response!  I've started over from scratch because I was confused.  The datasets did not match up to begin.  The National Land Cover Dataset was in Albers Equal Area projection, and the second dataset was recently created by someone out of LSU from remotely sensed technology.  It used Landsat TM images, and was projected to GRS 1980 UTM NAD 83 Zone 15 before classification.

Because my study area (the extent of my NLCD) covers both zones 15 & 16, and the second dataset was projected for Zone 15, I had to choose a reprojection for both datasets in order to merge the data later on.  I chose NAD 1983 State Plane Louisiana South (which covers all my study area along coastal Louisiana) as the reprojection for both.

I pulled both datasets into Arc, then used the Project Raster tool to reproject both datasets to the NAD 1983 State Plane Louisiana South.  After seeing the datasets still do not match up, I realized it was probably because the Data Frame projection responsible for visualizing the extent of the data on the screen did not match that of my datasets;  So I transformed the Data Frame projection to NAD 1983 State Plane Louisiana South, with the matching Datum D_North_American_1983.

However, the Data Frame properties indicates a projection of Lambert Conformal Conic as well, but the datasets do not specify such a projection.  Is this the reason both are still not matching up?  I'm still at a loss.  I appreciate your help!

Ashley

Frequent Contributor

ArcMap can project on the fly, so If the datasets did not line up correctly to begin with, then one of them was likely defined incorrectly. I would be less inclined to think that the NLCD data was incorrect, so the culprit is probably the recently created one. Have you tried defining the original dataset to zone 16?  Basically, if they didn't line up to begin with, then reprojecting them won't help. You need to find the correct original projection. Also, the data frame will have the projection of the first data set you add to your project. Sometimes it s better to start a new project and bring your dataset with the correct projecion in first to make them the same.

Occasional Contributor III

This is in addition to what Sephe has suggested so far.  Considering that the two images do not align (and assuming they should), one of them is not showing up in the correct place.  Have you tried comparing the images with some reference data to see which one is not aligned correctly?  This will help you figure out which one you need to investigate further to identify the underlying cause of the issue.  Two big possible causes are incorrect spatial reference definition (as pointed out by Sephe) or bad georeferencing.

New Contributor II

Thanks for the suggestion, guys. I pulled in a Louisiana shapefile,

converted it to a raster, and let that serve as the backdrop for a spatial

reference. After I pulled in both original datasets to start from

scratch, they matched up in the approximate location where they should. I

must have just kept reprojecting and lost site of what I had done!

Now, the NLCD is in Albers Conical Equal Area Projection, and the second

dataset is projected in NAD 1983 UTM Zone 15N. I've saved my work up until

this point in a new folder, and now I need to reproject both to a common

projection. Any recommendations on which projection to choose? My

research goal is to measure the extent to which restoring bald

cypress-water tupelo trees (identified in the 2nd dataset and categorized

as healthy, intermediate, and degraded) will attenuate waves from storm

surge; therefore, preserving the shape and area of the forest identified

int he second dataset is probably the most important thing to consider.

Thank you all so much!

Reply
0 Kudos
Frequent Contributor

Why don't you just project the second dataset into Albers Equal Area to match the NLDC?

Reply
0 Kudos
Occasional Contributor III

Albers, UTM, and Louisiana State Plane South are all in NAD83 so there are no transformations to worry about.  Given that your data crosses two UTM zones, this system should probably be avoided.  Limiting the choices to  those that have been discussed, either Albers or Louisiana State Plane South should work fine for your needs.  I am not a coordinate system expert but both of these systems are widely used for spatial analysis.  Given that your data would be closer to the origin of the Louisiana State Plane South system, it should be susceptible to less distortion versus a continental system like Albers, which is one of the reasons local coordinates systems are always being pushed.  BUT I am not an expert and cannot confirm that this logic always holds true or applies at all in this case.  Going as local as is reasonable is just a rule of thumb I try to follow (since I am no expert and need "rules of thumb").  Seems to me they are both based on very similar projections so I don't think you can go wrong here.  You could always do some testing by projecting the cypress dataset into each of the prospective systems and running area calculations to see how much the results differ.  If you are seeing large differences in area based on these tests then coordinate system selection may warrant further consideration.

Lambert Conformal Conic (used for state plane):  ArcGIS Help (10.2, 10.2.1, and 10.2.2)

Albers Equal Area Conic:  ArcGIS Help (10.2, 10.2.1, and 10.2.2)

View solution in original post

Frequent Contributor

Great idea, Gabe. My gut feeling is that difference won't be significant.