2

When creating a raster from a polygon (the CRS of the polygon is a geographic coordinate system), the raster is created in the projected coordinate reference system of the data frame, which is a projected one. I need to do so because I need to specify the Cell size of the raster in meters.

I also need to deliver this raster (previously converted to polygon) in WGS84 (GCS); I am considering to do it so: Changing the Data frame coordinate reference system to WGS84, and, on the result raster, do Data > Export Data and, for the option “Use the same coordinate system as:”, choose “the data frame”.

Is this an accurate way of dealing with that? I mean, aren’t there modifications in areas or coordinates of the delivered shape layer?

1

EDIT : When you reproject your raster there will be a resampling, so the new data will be degraded. It is therefore better to convert directly from a feature class with the output CRS that is most suited to your need, in other words you should reproject your feature class that is currently in WGS 84 into a projected CRS if you want your final data to be in WGS 84. The conversion of feature class will be exactly reversible if you don't change the datum, you just need to be aware that only the vertices are projected. therefore, if you have long segments without vertices, it is recommended to "[densify][1]" before your project.

The size of your pixels will affect the precision of your feature to raster conversion, as well as the precision of the resampling. It will not, however, affect the accuracy of the area of your polygon linked with projection (if your projection is not equal-area, the average over or under-estimation of the polygons area will not depend on the pixel size)

Now let us consider the choice of a CRS.

With a geographic (lat/long) coordinate system, the cell size will be expressed in degree. The cells of a raster of the same width and height everywhere, so the actual (on the ground) extent of grid cells will change depending on your latitude. If you want to know the size in meter of on cell in a lat/long system, you can have a good approximation with the knowledge that one degree is approximately 111.3 km at the equator, but that the width of your pixel has to be corrected by the cosinus of the latitude. Therefore, if you need a final raster in Lat/long, forget about an exact cell size in meter and use a convenient approximation based on your location (e.g. the minimum or the average of a degree of Lat and a degree of Long). For instance, at a latitude of 60°N, a square of one by one meter is approximately 8.9847e-6 degree in the Y direction and 1.797e-5 degree in the X direction.

With a projected coordinate system, the cell size is expressed in a cartesian system (with a unit in meter, in feet...) and this value is constant everywhere on the grid. Of course, you are in a projected coordinate system, so there is a small distortion due to the move from a sphere to a plane, but you should not be too much afraid of this distortion because you can usually find local projection that reduce the distortion below the error range of your measuring tool/method. Therefore, if you don't need to have a raster in Lat/long, use a (local) projected coordinate system where you project your feature class, then convert the pojected feature class to raster.

Remark : If you do not perform your analysis in spherical geometry, I recommend working in a projected CRS.

  • The problem is that I want to control the cell size of the output raster by assigning a value in meters. I am aware the best would be to convert directly from a feature with the output CRS (a geographic one), but... how can I assign the raster cell size in meters? – Charly Sep 22 '14 at 21:03
  • Anyway, @radouxju, won't it be better towards a less shape/area distortion of the final result to proceed in this way?: Create the raster from the shape which is in a Geographic CRS (WGS84), having set before the CRS of the Data Frame to a projected one (WGS84 UTM) in order to be able to give cell size in meters, which be a small value in relation with the actual required resolution in order to get accurate result when the expected resampling happens; then, once the raster is in a projected CRS, reproject it to WGS84 (the resampling is expected that involves a low shape/area distortion)? – Charly Sep 26 '14 at 9:17
  • Since I can create a raster in projected CRS, from WGS 1984 (the CRS of the feature class) by changing the Data Frame CRS to that projected one, I don't need to change any CRS but I can handle different units. I expect that you or somebody can validate the approach explained in my last comment. – Charly Sep 26 '14 at 19:12
  • I went to the chat room and read your message. I left a response there. Could you say me anything? – Charly Sep 30 '14 at 19:26
  • thank you. I don't quite understand all statements in your response. Anyway, since all this matter is about being able to set a specific cell size of rasters (that is, a cell size in meters or its equivalent in degrees), what about the next way around?: 1. set the Data Frame CRS (which originally is in WGS84) to a projected one; 2. Polygon to raster conversion, and setting the raster cell size in meters in the conversion tool window; then, the resulting raster (from now the “reference raster”) will be in a projected CRS; 3. Set the Data Frame CRS back to WGS84; – Charly Oct 3 '14 at 19:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.