Format conversion

Raster

Let's try converting a Sentinel 2 dataset to Cloud-optimized GeoTIFF (COG) using gdal raster convert

$ gdal raster convert S2B_MSIL2A_20260423T094029_N0512_R036_T34TDR_20260423T115714.SAFE/MTD_MSIL2A.xml TDR.tif --format COG

ERROR 6: COG driver does not support 0-band source raster

OK, we need to select one of the subdatasets:

$ gdal raster convert SENTINEL2_L2A:S2B_MSIL2A_20260423T094029_N0512_R036_T34TDR_20260423T115714.SAFE/MTD_MSIL2A.xml:10m:EPSG_32634 TDR.tif --format COG

Wait ~ 1 minute.

Let's check:

$ gdal info TDR.tif

[ ... snip ... ]
Image Structure Metadata:
  LAYOUT=COG
  COMPRESSION=LZW
  INTERLEAVE=PIXEL
[ ... snip ... ]
Band 1 Block=512x512 Type=UInt16, ColorInterp=Red
  Description = B4, central wavelength 665 nm
  Overviews: 5490x5490, 2745x2745, 1373x1373
[ ... snip ... ]

Exercise

1. Play with the different compression methods and observe effects on image size.

2. Try creating a 8-bit JPEG-compressed tiled GeoTIFF file with the red, green and blue bands.

   (hint)

   Hint

   Use gdal raster scale

3. Improve the visual result by reducing the effect of extreme pixel values (outliers) using statistical clamping.

   (hint)

   Hint

   Use gdal raster select

   You can improve the visual result by clamping the input range to 2 standard deviations of the mean.

==> Exercise solution for raster format conversion.

Vector

Convert the timisoara.osm.pbf to GeoPackage, with the other_tags field as JSON using gdal vector convert

$ gdal vector convert timisoara.osm.pbf timisoara.gpkg --open-option TAGS_FORMAT=JSON

Let's check:

$ gdal info timisoara.gpkg

INFO: Open of `timisoara.gpkg'
  using driver `GPKG' successful.

Layer name: points
Geometry: Point
Feature Count: 20415
Extent: (20.167133, 45.346126) - (22.818334, 46.318563)
Layer Coordinate Reference System:
  - name: WGS 84
  - ID: EPSG:4326
  - type: Geographic 2D
  - area of use: World, west -180.00, south -90.00, east 180.00, north 90.00
Data axis to CRS axis mapping: 2,1
FID Column = fid
Geometry Column = geom
osm_id: String (0.0)
name: String (0.0)
barrier: String (0.0)
highway: String (0.0)
ref: String (0.0)
address: String (0.0)
is_in: String (0.0)
place: String (0.0)
man_made: String (0.0)
other_tags: String(JSON) (0.0)

Layer name: lines
[ ... snip ... ]

Layer name: multilinestrings
[ ... snip ... ]

Layer name: multipolygons
[ ... snip ... ]

Layer name: other_relations
[ ... snip ... ]

Exercise

1. Convert timisoara.osm.pbf to GeoJSON

   (hint)

   Hint

   Use the --input-layer argument to generate one output file per input layer.

2. If you open the above timisoara.gpkg in QGIS, you can see that the other_relations layer cannot be displayed, because it contains geometry collection features that QGIS does not support.

   Try "exploding" those collections in single primitives (points, lines, polygons).

   (hint)

   Hint

   Use gdal vector explode-collections

3. Create a layer for each geometry type in the other_collections layer.

   (hint)

   Hint

   Use gdal vector filter and Spatialite ST_GeometryType function

==> Exercise solution for vector format conversion.