I have a LiDAR dataset with some missing data right in the middle of my map. I would like to use a 5m DEM to fill in the missing areas.

Is there a way in QGIS to merge the two but only use the 5m data in the missing location?

2 Answers 2


I'm not aware of any way to do it from within QGIS but GDAL can do it from the commandline. The basic steps are:

  • Crop, upsample, and smooth your low-res data to match your high-res data; you may also need to reproject the SRS.

    You can determine the appropriate parameters by running gdalinfo against your DEM files. SRC_SRS should match the value from your low-res file while DST_SRS, X_RES, and Y_RES should match the values from your high-res files. Note that the SRS specifications must be of the form ESPG:nnnn and not the textual form that gdalinfo outputs. The correct numeric form for a given projection can be looked up at https://epsg.io

    The coördinates given as arguments to -te (south-west and north-east corners in the projections units, at least for UTM) should be large enough to cover the area missing from the high-res data but the area covered shouldn't be much too large because the low-res DEM will be upsampled to the higher resolution.

    Smoothing the low-resolution data (e.g. with -r lanczos) is important; if omitted you'll end up with a blocky, checkerboard pattern in the low-res areas.
gdalwarp -s_srs SRC_SRS -t_srs DST_SRS -te EAST_0 NORTH_0 EAST_1 NORTH_1 -tr X_RES Y_RES -r lanczos SRC.img DST.tif
  • Merge the files into a single DEM:
gdal_merge.py -o OUT.tif DEM_FILES

(if not specified, the default output file is out.tif) DEM_FILES is a list of all the DEMs to combine. The lowest resolution files should be listed first and the highest-resolution should be listed last, per the man page:

In areas of overlap, the last image will be copied over earlier ones.

  • Process the resulting combined DEM as desired:
gdaldem hillshade OUT.tif OUTPUT_HILLSHADE.tif

For example:

  • imgn45w074_13.img is a ⅓ arc-second (10m) resolution, 1°×1° coverage, NAD-83 projection DEM from the USGS NED
  • 18TXQ601902.img and 18TWQ598902.img are 1m resolution, 1.5km×1.5km coverage, UTM 18N LIDAR DEMs (acquired from https://orthos.dhses.ny.gov/)
gdalwarp -s_srs EPSG:4269 -t_srs EPSG:26918 -te 598500.000 4902000.000 604500.000 4903500.000 -tr 1 -1 -r lanczos imgn45w074_13.img imgn45w074_13.utm_18N.tif
gdal_merge.py 18TXQ601902.img 18TWQ598902.img imgn45w074_13.utm_18N.tif
gdaldem hillshade out.tif Hillshade.tif
gdal_contour -i 1 -nln Crows out.tif Crows-1m/

Here's the resulting hillshade (reduced in resolution for the web) Hillshade (actual resolution is higher) From left to right, the first and third squares are derived from 1m LIDAR data and the remaining portion of the image is from the NED radar data. Here's a close-up: enter image description here Importing Crows-1m/Crows.shp and Hillshade.tif into QGIS this gives the following (close-up shown): enter image description here

I'm not sure how to handle the visible lines at the high-res/low-res boundaries. When I've encountered this in the past the solution was to combine all of the input files into a single VRT file. In this case, however, using a VRT file makes no difference.

Things are a bit trickier if you want to work with VRT files instead of using gdal_merge.py to create a new physical file. gdalbuiltvrt is very fussy about the input files being the same resolution, projection, etc. You'll need to do something like this:

gdalwarp -s_srs SRC_SRS -t_srs DST_SRS -te EAST_0 NORTH_0 EAST_1 NORTH_1 -tr X_RES Y_RES -r lanczos LOW_1.img DST_1.tif
gdalwarp HIGH_2.img DST_2.tif
gdalwarp HIGH_3.img DST_3.tif
gdalwarp HIGH_N.img DST_N.tif
gdalbuildvrt OUTPUT.img.vrt DST_1.tif DST_2.tif DST_3.tif DST_N.tif
  • This works great! I am still struggling to make the boundary between the high res/low res data less noticeable.
    – Andy G
    Commented Sep 1, 2017 at 13:17
  • @AndyG I'm not sure there's an easy solution to that. Looking at a the raw elevation data from the example I gave (just drag-and-dropping the TIFs/IMGs into GQIS) there's a visible line there too. Points along the boundary show up to about a 7m difference in elevation. It makes sense if you think about it; unless you're dealing with a flat surface you're going to have variation in elevation along the edge of the two datasets and that variation is going to show up as a line in the hillshades and contours. The only time the line disappears is when both the high-res and low-res datasets have Commented Dec 7, 2017 at 4:40
  • the same elevation. You can see this in this zoomed in section of the example above: alephnull.net/software/gis/Screenshot%20cropped.png Commented Dec 7, 2017 at 4:44
  • To eliminate this problem one would need to blend the two datasets along the edge. I don't know of a tool that will do this but if one really needed/wanted seamless edges it would not be too difficult to write a small program to do that. The basic technique would be: Leave the low-res data as is; the high-res data values should fade to the low res values as you approach the edge of the high-res data. Let me know if you'd like me to post some simple pseudo-code to do this. Commented Dec 7, 2017 at 4:50

You can do this from within the QGIS GUI using Processing -> GRASS -> r.patch (https://grass.osgeo.org/grass70/manuals/r.patch.html), which will fill in areas of no data in an input raster using data from another raster.

This does not do any form of smoothing though, so you will still end up with blocky edges as per the answer from @Alex Hajnal above.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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