Point value Change after projection - USGS DEM Raster VS Imagery Layer from Esri

[Jaslam]

Hello Tech.

I'm working on a project that requires National level DEM data, and I'd like to thank ESRI for providing the imagery layer (https://www.arcgis.com/home/item.html?id=0383ba18906149e3bd2a0975a0afdb8e).

1. For the most part, the extracted point values from the USGS raster and the ESRI imagery layer are very close. (Col 2 , Col 3 &Col 6)
2. However, when I project those raster’s (WGS 1984 Web Mercator Auxiliary Sphere), the point value difference between projected raster’s from USGS and ESRI were significantly high. (Col 4, Col 5 & Col 7)

Shape *	Raster exported from USGS	Raster exported from ESRI_layer	USGS_Projected           (WGS-1984)	ESRI_layer_Projected              (WGS-1984)	Abs(Raster exported        from USGS-Raster    exported from   ESRI_layer)	Abs(USGS_Projected (WGS-1984)-ESRI_layer_Projected (WGS-1984))
Point	3071.917	3071.913	3032.042	3071.913	0.004	39.871
Point	3268.988	3268.984	3297.935	3268.984	0.004	28.951
Point	3496.321	3496.318	3488.198	3516.118	0.003	27.92
Point	2501.627	2501.627	2527.583	2501.627	0	25.956
Point	3253.484	3253.491	3246.843	3271.352	0.007	24.509
Point	2932.126	2932.129	2932.126	2955.189	0.003	23.063
Point	3088.094	3088.087	3088.094	3110.907	0.007	22.813
Point	2760.626	2760.628	2789.526	2766.848	0.002	22.678
Point	2647.81	2647.816	2625.496	2647.816	0.006	22.32
Point	2546.791	2546.793	2568.551	2546.793	0.002	21.758
Point	2973.761	2973.768	2955.436	2976.748	0.007	21.312
Point	2899.172	2899.171	2910.918	2889.789	0.001	21.129
Point	1535.169	1535.16	1535.169	1514.62	0.009	20.549

I tried various geographical transformations and resampling methods but received no improvement. What could be the cause of this, and how should I proceed?

Looking forward to hearing from You

[BojanŠavrič · Accepted solution]

Hi @Jaslam, 

I might have an answer for you, but I need a bit more information before I write a proper reply. 

In which Projected/Geographic and Vertical Coordinate Systems is your Raster exported from USGS and Raster exported from ESRI_layer data? 

I assume your target VCS for both datasets is WGS 1984, WKID::115700. Is this correct?

Thanks,
Bojan

[Jaslam]

Thanks @BojanŠavrič 

Raster exported from ESRI layer -  XY =  NAD_1983 Z=None

Raster exported from USGS - XY =  NAD_1983 Z=None

Target Project coordinate system is WGS 1984.

For more info sharing data set:

https://vandalsuidaho-my.sharepoint.com/:f:/g/personal/mjaslam_uidaho_edu/EoPaEm7to6xNgLsBZbu0LIgBth...

Attached file info:

USGS_1_n45w114_20130911 - Raster Exported from USGS
USGS_Proj_NAD_1983_To_WGS_1984_5 - Projected Raster Exported from USGS
ESRI_shape - Raster Exported from ESRI (Clipping Geometry = ID2.shp)
ESRI_shape_ProjectRaster - Projected Raster Exported from ESRI (Clipping Geometry = ID2.shp)
ID2.shp - Lower Idaho state shape file

[BojanŠavrič · Accepted solution]

@Jaslam,

We look at your workflow and I might be able to provide you with an explanation. 

These differences are caused by a resampling technique to source the projected rasters. In your case, this would be the Nearest Neighbor method. Since your USGS_1_n45w114_20130911 and ESRI_shape have different extents and raster grids, when they are projected to WGS84 Web Mercator, their projected grids aligned differently with the source raster grids and the Nearest Neighbor method picks different pixel to source its raster value. E.g. take your point with 3071.913 height from ESRI_shape_ProjectRaster layer and check all nearest 9 pixels round that point in USGS_1_n45w114_20130911. One of the pixels will have the height of 3032.042 as it is in USGS_Proj_NAD_1983_To_WGS_1984_5 layer. Differences vary based on the elevation changes in the point’s neighborhood. Some are higher, some lower, and some are exactly equal. Similar effect, but with a bit smaller differences happens with other resampling techniques for the same reason.

Please note that the Nearest Neighbor method is inappropriate for elevation surfaces. For more details, please check Project Raster online documentation page.

I hope this helps,
Bojan