Project raster from GCS to UTM
Currently I'm working on terrain analysis project; it's main goal to prepare geomorphologic basis for further landscape mapping procedure support. As the main source of information on the first stages we use 90m DEMs. Before analytical tools (such as slope, aspect etc) applying we need to prepare the data, so it makes me ask for a piece of advice.
My question can be broken down into two parts:
1) To cover all the Ukraine territory with DEMs we downloaded several 5×5 degrees datasets. Further analysis will be carried out for smaller UTM projected training sites. What should we do first – to mosaic all the datasets and reproject them (but how? Ukraine territory lies in 4 UTM zones – from 34 to 37) or clip raster datasets from general mosaic according to UTM coordinates and reproject them separately (mosaicing them after we can increase distortion)?
2) To analyze the data we need to prepare it, i.e. reproject from original GCS_WGS_1984 to projected UTM (for example WGS_1984_UTM_Zone_35N). To do this we need “project raster” tool – how to use it properly:
- Geographic transformation is said to be optional if we have the same datum (in our case D_WGS_1984). Nevertheless, if we decided to specify parameters what should be written there?
- Which resampling technique fits with our needs best (we chose bilinear interpolation)?
- Output cell size probably must be set to 90m according to original raster resolution?
- The x and y coordinates used for pixel alignment (its 0;0 or 100;100 or …)?
Mapping Center Answer:
First, UTM is not an appropriate projected coordinate system for analysis of the entire study area--you'd have to pick a zone and then data in every other zone would be distorted, moreso for zones farther away from the zone you picked.
Your most important dataset will be a mosaic the DEMs using the native GCS (sounds like it is WGS84 which will be fine).
For smaller study areas you can clip raster datasets from that to do analyses, and then derive analytical surfaces that can be projected into the appropriate UTM zone.
The more challenging problem is choosing a coordinate system for analysis of the entire dataset. Given the types of surfaces you want to derive, e.g., slope, aspect, hillshade, etc. determine which of distance, shape, or direction distortions need to be minimized. Then choose a projected coordinate system that matches your requirement. For instance, if slope is your most useful derived surface, know that slope is based heavily on distance would need to be produced after projecting into projected coordinate system that minimizes distance distortion. That would allow you to validly compare slope values anywhere in the Ukraine.
Many projection algorithms introduce error and artifacts into the DEM and deriving analytical surfaces from those projected DEMs compounds that error. Thus, if you must project first, then you must either smooth (Focal Stats) or resample (e.g. bilinear) to reduce or eliminate the errors introduced by projecting. The trick is to evaluate whether your projection or the smoothing of projected result did more damage to the analytical information; in other words you need to know where your bias is.
If you would like to post a comment, please login.