Reproject + Snapping + Clipping

[AdamCinderich]

I am working with climate model output (6 models all at a resolution of 50km but in various projections).  I am also working with reanalysis data (at a resolution of 32km).

I have been able to bring in my NetCDF files as Raster Layers.  I have defined their projections and have even been able to successfully REproject all of them to a common projection (I chose Lambert Conic Conformal for North America as this is also the projection of my reanalysis data).  The problem I am having however, is that the raster pixels do not overlap completey (due to the fact that the domains of each climate model are different �?? and therefore the pixels originate at different origins).  I have reprojected, importing the extent of LCC from one model and using the �??snapping environment�?� to that common extent. This seems to have fixed the problem because now the pixels overlap, but I am not 100% sure that this is reliable.

My major question is this: does using the snapping environment �??shift�?� or move pixels right/left/up/down or does the interpolation actually reassign values to the upper corner based on the interpolated data?  If it shifts (my temperature data in this case) I can not use this when trying to subtract one map from another to see the differences between model output and observed data.

[anthonybaker]

The easiest way that I know of to align pixels is to composite the raster datasets together into a single raster with multiple bands.  The cell size can be determine as a least common denominator of the cell sizes (perhaps 10km in your case).  The bands can then be separated after the composite step.

Curious if the the netCDF data vector or raster data?

[RafaGutierrez]

I have been able to bring in my NetCDF files as Raster Layers.  I have defined their projections and have even been able to successfully REproject all of them to a common projection (I chose Lambert Conic Conformal for North America as this is also the projection of my reanalysis data).

I think I'm having the exact same problem. The data I'm working with is in a GCS where the origin is 20.E lon and ends around 405 long. My guess is that the folks who made the data don't need to analyze it in a common geographic projection and so they stuck with positive numbers for the longitude.

My question for you is what CRS did you define your data sets to? For me, I guess it's GCS but since it runs off the grid, it doesn't make sense to define it to a grid that's -180 to 180.

The problem I am having however, is that the raster pixels do not overlap completey (due to the fact that the domains of each climate model are different �?? and therefore the pixels originate at different origins).

I get a gap as well, when it's reprojected on the fly with other world data. Different domains for sure but the data are still there. And b/c it goes beyond 360 degrees, its as if it can't physically be displayed on itself where there is overlap.

Also, I'm confused about the workaround you used. Can you explain if you reprojected or let Arc reproject on the fly?

I have reprojected, importing the extent of LCC from one model and using the �??snapping environment�?� to that common extent. This seems to have fixed the problem because now the pixels overlap, but I am not 100% sure that this is reliable.

My major question is this: does using the snapping environment �??shift�?� or move pixels right/left/up/down or does the interpolation actually reassign values to the upper corner based on the interpolated data?  If it shifts (my temperature data in this case) I can not use this when trying to subtract one map from another to see the differences between model output and observed data.

I wonder since I have voids where there is land (global data set) I could actually see if there's a shift or just a resample. Giving it a shot now. Will let you know...