Raster processing

Eight runnable Hop pipelines for GeoTIFF, COG, clipping, resampling and building-height statistics.

A raster travels through Hop as a Raster value, not one row per pixel. Read a source, derive a clip or a reprojection, then write a file or calculate statistics. The project includes all six raster transforms: Reader, Info, Clip, Reproject, Writer and ZonalStats.

Download all eight pipelines and data

Set up the example project once

Install Spatial Hop, then extract the complete ZIP into a writable folder. Keep data, metadata, output and all pipelines together.
In Hop GUI, create a project named raster-processing. Set Home folder to the extracted raster-processing folder and use the included project-config.json. Use the existing configuration when prompted.
Select the project, open a pipeline and select the included local run configuration.
Run it. Every pipeline is independent. Only zonal-online.hpl needs internet; no database or separately installed GDAL is required.

All local paths use ${PROJECT_HOME}. Keep the supplied data unchanged. Overwrite is disabled: to repeat an example, choose a new output name or remove only its previously generated outputs. The polygon example creates one TIFF per building.

Choose a tutorial

Tutorial	Pipelines	Result
Read, inspect and write	`read-write-geotiff.hpl`, `read-write-cog.hpl`	GeoTIFF and COG with internal overviews.
Clip by bounding box	`clip-bbox.hpl`	A 250 × 250 m crop on the original grid.
Clip by building polygon	`clip-polygon.hpl`	33 separate, masked TIFFs.
Reproject and resample	`reproject-nearest.hpl`, `reproject-bilinear.hpl`	EPSG:3857 rasters with 1-unit pixels.
Building height statistics	`zonal.hpl`	33 building records with height statistics.
Statistics from an online COG	`zonal-online.hpl`	The same analysis using HTTP range reads.

Meet the sample data

Buildings only. data/buildings.gpkg, layer buildings, contains 33 polygons selected with art_txt = 'Gebaeude' from the MOpublic dataset 2549. Other land-cover classes are not included. The ordinary building_id attribute copies the original T_Id primary key so it remains available in Hop and output datasets. Original attributes, INTERLIS identifiers and coordinates are retained.

2019 nDSM. data/ndsm-2019.tif is a losslessly compressed crop of the normalised building surface model, not the 2023 raster. The crop adds a 10-metre margin around the buildings, rounded to source pixel boundaries. No values were resampled or reprojected.

Property	Value
CRS	EPSG:2056 — CH1903+ / LV95
Size	2322 × 4709 pixels, one band
Pixel size	0.25 × 0.25 metres
Extent	E 2611454.75–2612035.25; N 1233360.75–1234538.00
Numeric type / NoData	Float32 / `-9999`
Measurements	Normalised heights in metres above terrain

The current building inventory and the 2019 raster describe different dates. Four buildings have no valid raster samples in this snapshot. Their statistics remain null with status NO_VALID_PIXELS; they are not assigned a height of zero. Even a valid maximum is a sampled nDSM value inside a footprint, not an exact surveyed roof height.

Source: Kanton Solothurn, Amt für Geoinformation, MOpublic and LiDAR 2019. Source-data terms allow commercial and non-commercial use; attribution is recommended. The ZIP contains full provenance and checksums in data/README.md and data/provenance.json.

Tested setup and shared checks

Tested 3 October 2026, using Apache Hop 2.19.0, Temurin Java 21.0.10, macOS ARM64 and the installed spatial plugin builds. The ZIP’s tested-environment.json records their exact SHA-256 hashes. This is a fixed test record, separate from the website’s latest distribution download. Snapshot versions can share a name while containing different builds.

All eight pipelines were executed from an extracted project with a space in its path. GDAL 3.13.3, NumPy and Shapely independently checked output rasters, masks, CRS and statistics. Online and local statistics matched: 29 OK, 4 NO_VALID_PIXELS. Controlled fixtures also checked valid zero, NoData, polygon holes, outside zones, interpolation, existing-file protection and network failure.

Reader creates a source description. Clip and Reproject derive raster values; pixel evaluation happens at a consumer such as Writer or ZonalStats. Keep one writer copy per destination, avoid filename collisions and inspect the result rather than relying only on a green pipeline status.

Assigning a CRS is not reprojection

An explicit geometry CRS describes the coordinates already present. Raster Reproject actually transforms the grid. Pixel sizes are expressed in the target CRS units; EPSG:3857 uses map metres with latitude-dependent scale distortion.

Raster handbook (German) · Additional repository examples