New Zealand Elevation TIFF Specifications
April 21, 2026 · View on GitHub
LiDAR Capture Base Specification
Toitū Te Whenua Land Information New Zealand publishes a National Aerial LiDAR Base Specification which defines data quality requirements and some TIFF specifications that are also covered by the following data storage specification.
These include:
- GeoTIFFs with georeferencing information contained within each raster file. Our data processing workflows will also handle TFW World File sidecar file containing georeferencing information but this is not encouraged.
- Data provided in the NZTM2000 coordinate system and the NZVD2016 vertical datum.
- Void areas coded using "NODATA" value equal to -9999, identified in the appropriate location within the raster file header.
Cloud-Optimised GeoTIFFs and the specific compressions used for storage are not part of the base specification, and are outputs of LINZ data processing workflows.
Elevation Data Storage Specification
Applies to Digital Elevation Models, Digital Surface Models and Hillshades published in s3://nz-elevation after being processed by Toitū Te Whenua Land Information New Zealand.
For data processing scripts, see linz/topo-imagery.
For workflow configuration, see linz/topo-workflows.
File Format
| Property | Value |
|---|---|
| Container | GeoTIFF (.tiff) |
| Profile | Cloud-Optimised GeoTIFF (COG) |
| GDAL Driver | COG (-of COG) |
A summary of Cloud-Optimised GeoTIFF characteristics is provided in the GDAL docs.
Coordinate Reference Systems
| Type | Name | EPSG Code |
|---|---|---|
| Projection | New Zealand Transverse Mercator 2000 (NZTM2000) | 2193 |
| Vertical Datum | New Zealand Vertical Datum 2016 (NZVD2016) | 7839 |
Each TIFF contains the spatial reference information for NZTM2000. NZVD2016 is not recorded in the GeoTIFF metadata.
Spatial Resolution
1m for all LiDAR-sourced DEMs, DSMs and Hillshades.
s3://nz-elevation also contains a contour-derived DEM that is 8m spatial resolution.
Band Configuration
| Data Type | Bands | Bit Depth | Colour Interpretation | Nodata | Interleave |
|---|---|---|---|---|---|
| DEMs/DSMs | 1 (elevation) | float32 | Gray | -9999 | Band |
| Hillshade | 1 (greyscale) | uint8 | Gray | 0 | Band |
No alpha band is used for elevation data.
COG Driver Options
The following GDAL COG driver creation options are applied to all TIFFs:
| GDAL Creation Option | Value | Notes |
|---|---|---|
-co BLOCKSIZE | 512 | 512×512 pixel tiles (default) |
-co NUM_THREADS | ALL_CPUS | Use all available CPUs for compression, overview generation, etc |
-co SPARSE_OK | TRUE | Omit empty tiles (null byte) rather than writing blank data |
-co BIGTIFF | NO | Classic TIFF only (GDAL will raise an error if output exceeds 4 GB and we target approx 1 GB TIFF sizes via automated tiling) |
-co OVERVIEWS | IGNORE_EXISTING | Always regenerate overviews (never build from pre-existing lossy overviews) |
-co STATISTICS | YES | Compute and embed raster statistics (ensures that GIS software can immediately accurately display the data) |
Compression
DEMs/DSMs - LERC
| GDAL Creation Option | Value | Notes |
|---|---|---|
-co COMPRESS | LERC | LERC compression provides control over the maximum error per-pixel |
-co MAX_Z_ERROR | 0.001 | Maximum full-resolution error of 1mm |
-co MAX_Z_ERROR_OVERVIEW | 0.1 | Maximum overview error of 10cm |
-mo AREA_OR_POINT | Area | Forced to Area for consistency (input DEMs and DSMs vary) |
-a_nodata / --nodata | -9999 | Explicit nodata value set on all bands |
LERC is a near-lossless compression allowing a configurable per-pixel Z error (can also be completely lossless, but this removes one of the main advantages of LERC). The 1mm full-resolution error is well within the ~20cm vertical accuracy of the source LiDAR data. Overview error is relaxed to 10cm as overviews are used for visualisation only.
Hillshades - ZSTD
| GDAL Creation Option | Value | Notes |
|---|---|---|
-co COMPRESS | ZSTD | ZSTD compression is a high-performance, fast, lossless compression algorithm |
-co LEVEL | 17 | Highest compression level that produces a significantly smaller file size (at the cost of slower compression, which is acceptable for our use-cases) |
-co PREDICTOR | 2 | Horizontal differencing predictor |
-co OVERVIEW_COMPRESS | ZSTD | ZSTD compression for overviews as well |
-co OVERVIEW_RESAMPLING | LANCZOS | Lanczos resampling for overview generation |
File Naming Convention
Output files are named using the LINZ tile index grid name, e.g. BK27_10000_0101.tiff.
Individual 1m resolution LiDAR survey outputs are tiled to the NZ 1:10k Tile Index.
The 8m contour-derived DEM and combined national DEM, DSM and Hillshade datasets are tiled to the NZ 1:50k Tile Index.
GDAL Contributions
Toitū Te Whenua Land Information New Zealand has paid for the following contributions to GDAL:
-of COGdriver, now the de facto raster data file format for geospatial data in cloud storage-co MAX_Z_ERROR_OVERVIEWoption, saving us up to ~15% file size for elevation data with overviews
Along with ongoing maintenance sponsorship.
If you benefit from any of these tools, please also consider sponsoring GDAL.