FABDEM (Forest And Buildings removed Copernicus DEM) is a global elevation map that removes building and tree height biases from the Copernicus GLO 30 Digital Elevation Model (DEM). The data is available at 1 arc second grid spacing (approximately 30m at the equator) for the globe. The FABDEM dataset is licensed under a Creative Commons "CC BY-NC-SA 4.0" license. For commercial use queries, please contact fabdem@fathom.global This dataset is published in support of the paper "A 30 m global map of elevation with forests and buildings removed" published by IOP in Environmental Research Letters at https://dx.doi.org/10.1088/1748-9326/ac4d4f. UPDATE 14/03/2022 - Tile N00E011_FABDEM_V1-0.tif was corrupted and has now been replaced. This has been reflected in the geotiff tags with the following text "NOTE=This file is a replacement for originally corrupted file for tile N00E011" -mo "UPDATED=2022-02-23"" UPDATE 18/01/2023 - A new version of this dataset is available as FABDEM V1-2 at https://doi.org/10.5523/bris.s5hqmjcdj8yo2ibzi9b4ew3sn Complete download (zip, 462.3 GiB)

FABDEM V1.2 STAC Catalog

  Dataset Description

FABDEM (Forest And Buildings removed Copernicus DEM) is a global elevation dataset that provides a comprehensive 30-meter resolution digital elevation model with building and tree height biases systematically removed from the Copernicus GLO 30 Digital Elevation Model (DEM). This enhanced dataset offers more accurate representation of bare-earth topography for hydrological modeling, flood risk assessment, and other geospatial… See the full description on the dataset page: https://huggingface.co/datasets/links-ads/fabdem-v12.

Title:

FABDEM V1-0 adjusted for the Ayeyarwady Delta in Myanmar by local spot height data from topographic maps

Citation:

Seeger, K., Minderhoud, P. S. J., Peffeköver, A., Vogel, A., Brückner, H., Kraas, F., Nay Win Oo, Brill, D. (2023): FABDEM V1-0 adjusted for the Ayeyarwady Delta in Myanmar by local spot height data from topographic maps. Zenodo, https://doi.org/10.5281/zenodo.7875856.

Supplement to:

Seeger, K., Minderhoud, P. S. J., Peffeköver, A., Vogel, A., Brückner, H., Kraas, F., Nay Win Oo, and Brill, D. (2023): Assessing land elevation in the Ayeyarwady Delta (Myanmar) and its relevance for studying sea level rise and delta flooding. EGUsphere [preprint], https://doi.org/10.5194/egusphere-2022-1425.

Abstract:

This digital elevation model is a version of the FABDEM V1-0 of Hawker et al. (2022; https://doi.org/10.1088/1748-9326/ac4d4f) that was adjusted for the Ayeyarwady Delta in Myanmar by local spot height data from topographic maps (scale 1:50,000) published in 2014 while source data was compiled between 2000 and 2004. The FABDEM V1-0 (Laurence Hawker, Jeffrey Neal (2021): FABDEM V1-0. https://doi.org/10.5523/bris.25wfy0f9ukoge2gs7a5mqpq2j7; CC BY-NC-SA 4.0) was projected to the Myanmar 2000 datum and clipped to the Ayeyarwady Delta region of interest. The vertical reference of the FABDEM V1-0 was transformed to EGM96 before applying a conversion to continuous mean sea level based on mean dynamic topography data (CNES-CLS18 dataset of Mulet et al. (2021; https://doi.org/10.5194/os-17-789-2021) that we transposed to EGM96). Subsequently, inland water bodies were masked using the water body mask of the Copernicus DEM (Airbus Defence and Space, 2020: Copernicus Digital Elevation Model Product Handbook Version 3.0, Airbus, 38 pp.) and cell values with an elevation of more than 7 m below mean sea level were removed.

From the topographic maps, the local spot heights outside of areas masked in the AD-DEM (Seeger et al. (2023): Local digital elevation model for the Ayeyarwady Delta in Myanmar (AD-DEM) derived from digitised spot and contour heights of topographic maps. Doi; CC-BY 4.0) were subtracted from elevation values of the FABDEM V1-0 at the same locations (~3630 spot heights). Empirical Bayesian Kriging with empirical data transformation and exponential modelling was applied to interpolate the height residuals and export the raster data at ~30 m grid cell resolution. The mask layer of the AD-DEM was applied to the height residual raster in order to correct for interpolations in areas of data paucity. Subsequently, the interpolated height residuals were subtracted from the pre-processed FABDEM. In delta areas outside the masked regions of the height residual raster, the elevation of the pre-processed FABDEM was maintained (see the figure in the uploaded dataset).

For further information on processing of local and global elevation data for the Ayeyarwady Delta in Myanmar, including DEM interpolation, determination of local mean sea level and vertical datum conversions, as well as DEM performance, see the corresponding paper and supplementary material.

File name: FABDEM_EGM96_MDT_AD_MMR2000_maskedCop_min7_adjusted_AD.tif

File format: GEOTIFF file

Spatial reference: MMR2000_46N

Vertical reference: local continuous mean sea level, i.e., mean dynamic topography (CNES-CLS18 dataset of Mulet et al. (2021; https://doi.org/10.5194/os-17-789-2021) transposed to EGM96

Cell size: 30 × 30 m

This collection provides a seamlessly merged, hydrologically robust Digital Elevation Model (DEM) for the Murray Darling Basin (MDB), Australia, at 5 m and 25 m grid cell resolution.

This composite DEM has been created from all the publicly available high resolution DEMs in the Geoscience Australia (GA) elevation data portal Elvis (https://elevation.fsdf.org.au/) as at November 2022. The input DEMs, also sometimes referred to as digital terrain models (DTMs), are bare-earth products which represent the ground surface with buildings and vegetation removed. The DEMs were either from lidar (0.5 to 2 m resolution) or photogrammetry (5 m resolution) and totalled 852 individual DEMs.

The merging process involved ranking the DEMs, pairing the DEMs with overlaps, and adjusting and smoothing the elevations of the lower ranked DEM to make the edge elevations compatible with the higher-ranked DEM. This method is adapted from Gallant 2019 with modifications to work with hundreds of DEMs and have a variable number of gaussian smoothing steps.

Where there were gaps in the high-resolution DEM extents, the Forests and Buildings removed DEM (FABDEM; Hawker et al. 2022), a bare-earth radar-derived, 1 arc-second resolution global elevation model was used as the underlying base DEM. FABDEM is based on the Copernicus global digital surface model.

Additionally, hillshade datasets created from both the 5 m and 25 m DEMs are provided.

Note: the FABDEM dataset is available publicly for non-commercial purposes and consequently the data files available with this Collection are also available with a Creative Commons NonCommercial ShareAlike 4.0 Licence (CC BY-NC-SA 4.0). See https://data.bris.ac.uk/datasets/25wfy0f9ukoge2gs7a5mqpq2j7/license.txt Lineage: For a more detailed lineage see the supporting document Composite_MDB_DEM_Lineage.

DATA SOURCES 1. Geoscience Australia elevation data (https://elevation.fsdf.org.au/) via Amazon Web Service s3 bucket. Of the 852 digital elevation models (DEMs) from the GA elevation data portal, 601 DEMs were from lidar and 251 were from photogrammetry. The latest date of download was Nov 2022. The oldest input DEM was from 2008 and the newest from 2022.

1. FABDEM - Forests and buildings removed DEM based on the 1 arc-second Copernicus global digital surface model. Hawker, L., Uhe, P., Paulo, L., Sosa, J., Savage, J., Sampson, C., Neal, J., 2022. A 30 m global map of elevation with forests and buildings removed. Environ. Res. Lett. 17, 024016. https://doi.org/10.1088/1748-9326/ac4d4f

METHODS Part I. Preprocessing The input DEMs were prepared for merging with the following steps: 1. Metadata for all input DEMs was collated in a single file and the DEMs were ranked from finest resolution/newest to coarsest resolution/oldest 2. Tiled input DEMs were combined into single files 3. Input DEMs were reprojected to GA LCC conformal conic projection (EPSG:7845) and bilinearly resampled to 5 m 4. Input DEMs were shifted vertically to the Australian Vertical Working Surface (AVWS; EPSG:9458) 5. The input DEMs were stacked (without any merging and/or smoothing at DEM edges) based on rank so that higher ranking DEMs preceded the lower ranking DEMs, i.e. the elevation value in a grid cell came from the highest rank DEM which had a value in that cell 6. An index raster dataset was produced, where the value assigned to each grid cell was the rank of the DEM which contributed the elevation value to the stacked DEM (see Collection Files - Index_5m_resolution) 7. A metadata file describing each input dataset was linked to the index dataset via the rank attribute (see Collection Files - Metadata)

Vertical height reference surface https://icsm.gov.au/australian-vertical-working-surface

Part II. DEM Merging The method for seamlessly merging DEMs to create a composite dataset is based on Gallant 2019, with modifications to work with hundreds of input DEMs. Within DEM pairs, the elevations of the lower ranked DEM are adjusted and smoothed to make the edge elevations compatible with the higher-ranked DEM. Processing was on the CSIRO Earth Analytics and Science Innovation (EASI) platform. Code was written in python and dask was used for task scheduling.

Part III. Postprocessing 1. A minor correction was made to the 5 m composite DEM in southern Queensland to replace some erroneous elevation values (-8000 m a.s.l.) with the nearest values from the surrounding grid cells 2. A 25 m version of the composite DEM was created by aggregating the 5m DEM, using a 5 x 5 grid cell window and calculating the mean elevation 3. Hillshade datasets were produced for the 5 m and 25 m DEMs using python code from https://github.com/UP-RS-ESP/DEM-Consistency-Metrics

Part IV. Validation Six validation areas were selected across the MDB for qualitative checking of the output at input dataset boundaries. The hillshade datasets were used to look for linear artefacts. Flow direction and flow accumulation rasters and drainage lines were derived from the stacked DEM (step 5 in preprocessing) and the post-merge composite DEM. These were compared to determine whether the merging process had introduced additional errors.

OUTPUTS 1. seamlessly merged composite DEMs at 5 m and 25 m resolutions (geotiff) 2. hillshade datasets for the 5 m and 25 m DEMs (geotiff) 3. index raster dataset at 5 m resolution (geotiff) 4. metadata file containing input dataset information and rank (the rank column values link to the index raster dataset values) 5. figure showing a map of the index dataset and 5m composite DEM (jpeg)

DATA QUALITY STATEMENT Note that we did not attempt to improve the quality of the input DEMs, they were not corrected prior to merging and any errors will be retained in the composite DEM.

The datasets includes processed ICESat-2 ATL03 processed elevation data, together with Sentinel-2 MSI bands (B2, B3, B4 and B8) that are used in a ANN training to correct FabDEM.

Additionally, examples of flood inundation maps are included, for simulations using FabDEM and the corrected ANN DEM. The inundations maps are compared with observed data to assess their performance.

The data covers a 50 km floodplain in the Upstream Yellow River

ICESat-2 segments were used to compare four known global and one new national DEM datasets.

Introduction

This dataset's snow depth data was derived using elevation differencing, which is simply the snow surface elevation (ICESat-2 ATL08) minus the reference surface elevation (obtained from Digital Elevation Models):

1. DEM Co-registration: DEMs are co-registered to ICESat-2 ATL08 snow-off reference without vertical bias adjustment.
2. Elevation Bias Correction: The elevation bias between the DEMs and ICESat-2 is corrected using ICESat-2 ATL08 snow-off segments.
3. Snow Depth Calculation: Determining snow depth by subtracting the bias-free reference ground elevation(from Step 2) from ICESat-2 ATL08 snow-on segments.

This dataset is presented in a tabular format, which simplifies the preprocess for machine learning models. While co-registration has been done (1), users have the flexibility to train a bias correction model again (2) and retrieve snow depth measurements anew (3). Alternatively, the snow depth can be directly used for various analytical purposes. Detailed methodologies for the co-registration, bias correction, and snow depth determination are thoroughly documented in the paper (under submission) to support users in leveraging this dataset for their research needs.

Meta Information

• Study Area: Mainland Norway
• Acquisition Period (ICESat-2): October 2018 to October 2020
• ICESat-2 data source: ATL08 (level3, version 5)
• Reference DEMs: Norway DTM1, Norway DTM10, Copernicus GLO30, FABDEM. (see reference links)
• Reference snow depth: ERA5 Land (hourly), ERA5 Land (monthly).
• Snow condition: The dataset contains snow depth retrieved (snow_on_alt08_segments_and_snow_depth.csv) and snow-free observations (snow_free_alt08_segments_and_dems.csv).
• Data Cleaning: No, this is a raw dataset that may contain outliers.
• Mask: Excluded water surface and permanent ice at a spatial resolution of 100 m.

Description

This dataset encapsulates a wide array of attributes derived from ICESat-2 observations, alongside measurements pertinent to snow depth, terrain, and environmental conditions across Mainland Norway. For detailed attribute descriptions, refer to the ICESat-2 ATL08 documentation. The dataset is structured into several columns, each representing a specific attribute:

1. 'latitude': Latitude coordinates of the data points in WGS 84.
2. 'longitude': Longitude coordinates of the data points in WGS 84.
3. 'segment_landcover': Land cover classification for each segment.
4. 'segment_snowcover': Snow cover classification for each segment.
5. 'h_te_best_fit': Best-fit elevation of the terrain.
6. 'h_te_std': Standard deviation of terrain elevation.
7. 'n_te_photons': Number of photons used for terrain elevation estimation.
8. 'subset_te_flag': Quality flag (5 = all geosegments available, 4 = four geosegments...).
9. 'segment_cover': Woody vegetation fractional cover derived from the 2019 Copernicus 100m shrub and forest fractional cover data product.
10. 'h_canopy': Canopy height above terrain from ICESat-2 (only for snow-off segments).
11. 'h_mean_canopy': Mean canopy height ICESat-2 (only for snow-off segments).
12. 'canopy_openness': Canopy openness from ICESat-2 (only for snow-off segments).
13. 'h_canopy_winter': Canopy height above terrain from ICESat-2 (only for snow-on segments).
14. 'h_mean_canopy_winter':Canopy mean height from ICESat-2 (only for snow-on segments).
15. 'canopy_openness_winter':Canopy openness from ICESat-2 (only for snow-on segments).
16. 'tree_presence': the presence of trees in the segment (1 = tree, 0 = no tree, binary of h_canopy).
17. 'pair': Pair flag for ICESat-2.
18. 'beam': Beam flag for ICESat-2.
19. 'p_b': Pair and beam flag for ICESat-2.
20. 'region': Region identifier for ICESat-2.
21. 'cloud_flag_atm': Atmospheric cloud flag for ICESat-2.
22. 'urban_flag': Urban area flag for ICESat-2.
23. 'h_te_skew': Skewness of terrain elevation of segments.
24. 'snr': Signal-to-noise ratio for ICESat-2.
25. 'terrain_slope': Slope of the terrain from ICESat-2.
26. 'h_te_uncertainty': Uncertainty in terrain elevation estimation.
27. 'night_flag': Flag indicating nighttime data.
28. 'brightness_flag': Brightness flag for ICESat-2.
29. 'h_te_interp': Interpolated terrain elevation.
30. 'E': Easting coordinate in EPSG 32633.
31. 'N': Northing coordinate in EPSG 32633.
32. 'slope': Terrain slope computed from DTM10.
33. 'aspect': Terrain aspect computed from DTM10.
34. 'planc': Plan curvature computed from DTM10.
35. 'profc': Profile curvature computed from DTM10.
36. 'curvature': Overall terrain curvature computed from DTM10.
37. 'tpi': Terrain Position Index computed from DTM10.
38. 'tpi_9': TPI with a 90-meter radius.
39. 'tpi_27': TPI with a 270-meter radius.
40. 'wf_positive': Positive wind aspect index.
41. 'wf_negative': Negative wind aspect index.
42. 'smlt_acc': Snowmelt accumulation calculated from ERA5 Land monthly snow melting (currently not in use).
43. 'sf_acc': Snowfall accumulation calculated from ERA5 Land monthly snowfall (currently not in use).
44. 'sd_era': Snow depth from ERA5 Land reanalysis, coupled with ICESat-2 measurements at daily resolution,
45. 'sde_era': Snow depth linear interpolated from ERA5 Land reanalysis.
46. 'date': Date of data acquisition.
47. 'date_': Date in Pandas Datatime data dype.
48. 'month': Month of data acquisition.
49. 'difference': The elevation difference between segment and subsegment at the midpoint ( 'h_te_best_fit_20m_2' minus 'h_te_best_fit'). If you want to use h_te_best_fit_20m_2 instead of h_te_best_fit as elevation from ICESat-2, you can do it by df_after_dtm1 - difference, snowdepth_dtm1 - difference.

Columns on elevation difference and snow depth (in meters):

1. 'dh_after_dtm1': The elevation difference between the snow-free segment and DTM1 (ICESat-2 minus DTM1). This serves as an independent variable y in the bias correction model for DTM1. Here, 'after' means after co-registration.
2. 'snowdepth_dtm1': The elevation difference between the snow-on segment and DTM1 (ICESat-2 minus DTM1), representing the raw snow depth as measured against DTM1.
3. 'sd_correct_dtm1': Corrected snow depth using DTM1, adjusted by bias correction model.
4. 'df_dtm1_era5': Difference betwen 'sd_correct_dtm1' and 'sde_era'. (sd_correct_dtm1 minus sde_era), providing a comparison between corrected snow depth from DTM1 and snow depth from ERA5 Land reanalysis
5. 'dh_after_dtm10': The elevation difference between the snow-free segment and DTM10 (ICESat-2 minus DTM10), used in bias correction for DTM10.
6. 'snowdepth_dtm10': The elevation difference between the snow-on segment and DTM10 (ICESat-2 minus DTM10).
7. 'sd_correct_dtm10': Corrected snow depth using DTM10, adjusted by bias correction model.
8. 'df_dtm10_era5': Difference between 'sd_correct_dtm10' and 'sde_era'.
9. 'dh_after_cop30': The elevation difference between the snow-free segment and Copernicus GLO30 (ICESat-2 minus Copernicus GLO30).
10. 'snowdepth_cop30': The elevation difference between the snow-on segment and Copernicus GLO30.
11. 'sd_correct_cop30': The adjusted snow depth using Copernicus GLO30, adjusted by bias correction model.
12. 'df_cop30_era5': The discrepancy between 'sd_correct_cop30' and 'sde_era'.
13. 'dh_after_fab': The elevation difference between the snow-free segment and FABDEM (ICESat-2 minus FABDEM), used in bias correction for FABDEM.
14. 'snowdepth_fab': The elevation difference between the snow-on segment and FABDEM, representing the uncorrected snow depth.
15. 'sd_correct_fab': The corrected snow depth using FABDEM, adjusted by bias correction model.
16. 'df_fab_era5': The difference between 'sd_correct_fab' and 'sde_era'.

More explanation (especially on how the parameters are calculated, such as wind aspect index) is available in related works and blog posts on snow depth, and DEM bias correction.

This dataset includes a comprehensive collection of snow depth data and correlated environmental variables for Mainland Norway. Researchers can use this dataset to investigate the following:

• The difference between ICESat-2 and DEMs. For example, how 'df_after_dtm1' relates to terrain parameters.
• The residual bias of ICESat-2 derived snow depth, for example, snowdepth_dtm1 and bias-corrected sd_correct_dtm1. You can train a better bias correction to retrieve snow depth again. You can compare your model with my model by 'dh_reg_dtm1', 'dh_reg_dtm10', 'dh_reg_cop30', and 'dh_reg_fab', which are the elevation differences after bias correction for each DEM.
• The difference between ICESat-2-derived snow depth and snow depth from ERA5 Land, for example, 'df_dtm1_era5'.
• The spatial distribution of snow depth or subgrid variability.

Digital elevation model at 10 m resolution free of modern interferences of the hinterland of Ravenna, Italy.

This dataset includes:

Ravenna_hinterland_LocalDEM+.tif = digital elevation model at 10 m resolution free of modern interferences of the hinterland of Ravenna

Ravenna_hinterland_LocalDEM+.gpkg = vector point file containing the elevation data used to model the LocalDEM+ of the hinterland of Ravenna

The LocalDEM+ for the hinterland of Ravenna was created by interpolating ground control points (GCP) manually recorded by the Emilia-Romagna region [1]. In December 2018, the downloaded dataset still included modern artefacts among the almost 400.000 points, such as those over artificial fluvial banks, streets, railroads, and the A14 highway. Therefore, all elevation points not classified as “isolato al suolo” (= recorded on the bare soil) were removed to filter out these modern disturbances. Further manual cleaning was carried out to remove additional points along artificial infrastructures such as fluvial banks and the highway.

The original LocalDEM was created interpolating the resulting 160.481 points using the Inverse Distance Weighting method [2] creating a DTM at 10 m resolution devoid of modern interferences (Version v1).

The same 160.481 points dataset has also been interpolated via co-kriging [3] using FABDEM V1-2 [4] as the second correlated variable. Tiles “N044E011” and “N044E012” [5] were merged together before interpolation and 250.000 locations were randomly sampled within the merged grid extension using the “Random points” tool. Elevation for all 250.000 locations was extracted using the "Sample raster values” tool.

The improved LocalDEM+ shares the same spatial dimension and pixel resolution of 10 m, being as well devoid of modern interferences (Version v2).

[1] Data was retrieved from https://geoportale.regione.emilia-romagna.it/download/download-data on 21st December 2018 (last accessed on 10 February 2021).

[2] Mitas, L., & Mitasova, H. (2005). Spatial Interpolation. In P. Longley, M. F. Goodchild, D. J. Maguire, & D. W. Rhind (Eds.), Geographical Information Systems: Principles, Techniques, Management and Applications (Second Edition., pp. 481–832). Wiley.

[3] Gooverts, P. (1998). “Ordinary cokriging revisited”, Mathematical Geology 30, pp. 21–42, https://doi.org/10.1023/A:1021757104135

[4] FABDEM (Forest And Buildings removed Copernicus DEM) is a global elevation map that removes building and tree height biases from the Copernicus GLO 30 Digital Elevation Model (DEM). The data is available at 1 arc second grid spacing (approximately 30m at the equator) for the globe. FABDEM is introduced in Hawker et al. (2022), “A 30 m global map of elevation with forests and buildings removed”, Environmental Research Letters 17(2), https://dx.doi.org/10.1088/1748-9326/ac4d4f

[5] Data was retrieved from https://data.bris.ac.uk/data/dataset/s5hqmjcdj8yo2ibzi9b4ew3sn on 12th January 2024 (last accessed on 12th January 2024).

This database supports the work of the Digital Elevation Model Intercomparison eXperiment (DEMIX) working group (Strobl and others, 2021; Guth and others, 2021; Bielski and others, 2023, 2024). The two files have the database in CSV format, and a metadata file describing the contents of each field in the database.

To understand the use of the database, see the prepint (Bielski and others, 2023).

Changes to version 2 which is the only version you should use:

1. Added 2 new areas, Stateline and Canary Islands East which should have minimal differences between the DSM and the DTM and no significant changes over the last 20 years.

2. Added the country to the database

3. Added a number of areas in France

4. Added some additional tiles for a few areas

5. Total number of tiles almost doubled

6. Now using GDAL to compute the datum shift, horizontal and vertical, for USGS 3DEP

7. Fixed some anomalies computing pixel-is-area DEMs

8. Recomputed all the reference data and the version 1.0 GIS database (Guth, 2022)

9. New file naming conventions

References:

Bielski, C.; López-Vázquez, C.; Guth. P.L.; Grohmann, C.H. and the TMSG DEMIX Working Group, 2023. DEMIX Wine Contest Method Ranks ALOS AW3D30, COPDEM, and FABDEM as Top 1” Global DEMs: https://arxiv.org/pdf/2302.08425.pdf

Bielski, C.; López-Vázquez, C.; Grohmann, C.H.; Guth. P.L.; Hawker, L.; Gesch, D.; Trevisani, S.; Herrera-Cruz, V.; Riazanoff, S.; Corseaux, A.; Reuter, H.; Strobl, P., 2024. Novel approach for ranking DEMs: Copernicus DEM improves one arc second open global topography. IEEE Transactions on Geoscience & Remote Sensing. vol. 62, pp. 1-22, 2024, Art no. 4503922, https://doi.org/10.1109/TGRS.2024.3368015

Guth, P.L.; Van Niekerk, A.; Grohmann, C.H.; Muller, J.-P.; Hawker, L.; Florinsky, I.V.; Gesch, D.; Reuter, H.I.; Herrera-Cruz, V.; Riazanoff, S.; López-Vázquez, C.; Carabajal, C.C.; Albinet, C.; Strobl, P. Digital Elevation Models: Terminology and Definitions. Remote Sens. 2021, 13, 3581. https://doi.org/10.3390/rs13183581

Strobl, P.A.; Bielski, C.; Guth, P.L.; Grohmann, C.H.; Muller, J.P.; López-Vázquez, C.; Gesch, D.B.; Amatulli, G.; Riazanoff, S.; Carabajal, C. The Digital Elevation Model Intercomparison eXperiment DEMIX, a community based approach at global DEM benchmarking. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2021, XLIII-B4-2021, 395–400. https://doi.org/10.5194/isprs-archives-XLIII-B4-2021-395-2021

These are the data used for the analysis of globally derived synthetic rating curves (SRCs) in the Journal of Flood Risk Management study.

The 'Yellowstone_Basin' folder contains the data used to analyze SRCs at 29 stream gages throughout the Yellowstone basin.

The 'Yellowstone_2022_Floods' folder contains the data used to analyze the 2022 Yellowstone floods with SRCs and U. S. Geological Survey (USGS) high water marks.

All digital elevation model (DEM) data used are originally from the Forest and Buildings removed Copernicus DEM (FABDEM) version 1.2 (https://data.bris.ac.uk/data/dataset/s5hqmjcdj8yo2ibzi9b4ew3sn).

All synthetic streamflow data are from the GEOGLOW ECWMF Streamflow Service Version 2 (https://geoglows.ecmwf.int/).

USGS streamflow data were accessed using the Python library dataretrieval-python (https://github.com/DOI-USGS/dataretrieval-python).

Land cover data were accessed using the European Space Agency (ESA) WorldCover applied programming interface (API) (https://esa-worldcover.org/en/data-access).

This reference data supports the work of the Digital Elevation Model Intercomparison eXercise (DEMIX) working group (Strobl and others, 2021; Guth and others, 2021; Bielski and others,2023).

The data set consists of:

• Excel metadata file with two tabs:
  • Credits for each of 19 areas for the source of the high resolution data sets, the number of DEMIX tiles (Guth and others, 2023) in that area, the resolution of the source data, and credits for the processing.
  • The national mapping agencies (or equivalent) that provided the free high resolution data, and the files used for the vertical datum transformations.
• ZIP file with the 1” grids used to support the DEMIX work and create the DEMIX database (Guth, 2022). The file also includes databases for each area required to use the data in MICRODEM.

To understand the use of the reference data, see the forthcoming paper (Bielski and others, in prep).

References:

Bielski, C.; López-Vázquez, C.; Guth. P.L.; Grohmann, C.H. and the TMSG DEMIX Working Group, 2023. DEMIX Wine Contest Method Ranks ALOS AW3D30, COPDEM, and FABDEM as Top 1” Global DEMs: https://arxiv.org/pdf/2302.08425.pdf

Guth, Peter L. (2022). DEMIX GIS Database (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7402618 .

Guth, P.L., Strobl, P., Gross, K., & Riazanoff, S. (2023). DEMIX 10k Tile Data Set (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7504791

Guth, P.L.; Van Niekerk, A.; Grohmann, C.H.; Muller, J.-P.; Hawker, L.; Florinsky, I.V.; Gesch, D.; Reuter, H.I.; Herrera-Cruz, V.; Riazanoff, S.; López-Vázquez, C.; Carabajal, C.C.; Albinet, C.; Strobl, P. Digital Elevation Models: Terminology and Definitions. Remote Sens. 2021, 13, 3581. https://doi.org/10.3390/rs13183581

Strobl, P.A.; Bielski, C.; Guth, P.L.; Grohmann, C.H.; Muller, J.P.; López-Vázquez, C.; Gesch, D.B.; Amatulli, G.; Riazanoff, S.; Carabajal, C. The Digital Elevation Model Intercomparison eXperiment DEMIX, a community based approach at global DEM benchmarking. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2021, XLIII-B4-2021, 395–400. https://doi.org/10.5194/isprs-archives-XLIII-B4-2021-395-2021

This was developed for a forthcoming paper. A reference will be posted here when it is published.

This database supports the work of the Digital Elevation Model Intercomparison eXperiment (DEMIX) working group (Strobl and others, 2021; Guth and others, 2021; Bielski and others, 2024). The four files have the database tables in CSV format.

• Difference distributions for elevation, slope, and surface roughness. The provides continuity with \cite{BielskiOthers2024, GuthOthers2024}; for readers who want, it has the statistics like RMSE and LE90 for elevation and two LSPs, as well as the signed mean and median differences.
• FUV for a mixed suite of LSPs chosen to sample the full range of LSPs calculated from DEMs. These provide a better rankings of the test DEMs, and provides an estimate of the robustness of LSPs and suggest that some should be used with caution.
• FUV for the partial derivatives used for slope, aspect, and curvature.
• FUV for the suite of integrated curvature measures (Minar and others, 2020.

This version adds to CopDEM, ALOS AW3D30, and FABDEM:

• FathomDEM (Uhe and others, 2025)
• EDTM 1.1 (https://s3.eu-central-1.wasabisys.com/openlandmap/dtm/dtm.bareearth_ensemble_p10_30m_s_2018_go_epsg4326_v20230221.tif, https://doi.org/10.5281/zenodo.7676373)
• GEDTM 1.2—this is the best version of 1.2 (there are at least two others,which have no scaling for the elevations) (https://s3.opengeohub.org/global/dtm/v1.2/gedtm_rf_m_30m_s_20060101_20151231_go_epsg.4326.3855_v1.2.tif
• GEDTM 1.1—this has 0.9” spacing https://s3.opengeohub.org/global/edtm/gedtm_rf_m_30m_s_20060101_20151231_go_epsg.4326.3855_v20250611.tif, https://doi.org/10.5281/zenodo.15689805
• GEDTM v 0—this has 0.9” spacing https://s3.opengeohub.org/global/edtm/legendtm_rf_30m_m_s_20000101_20231231_go_epsg.4326_v20250130.tif, https://doi.org/10.5281/zenodo.14900181

The database contains 1381 tiles, about 10x10 km, in 140 areas. The tiles are based on the local projected grid, a change from earlier versions of the DEMIX database which used geographic outlines.

It does not consider the low altitude coastal DEMs; for those use version 3 (https://zenodo.org/records/13331458 ).

References:

Bielski, C.; López-Vázquez, C.; Grohmann, C.H.; Guth. P.L.; Hawker, L.; Gesch, D.; Trevisani, S.; Herrera-Cruz, V.; Riazanoff, S.; Corseaux, A.; Reuter, H.; Strobl, P., 2024. Novel approach for ranking DEMs: Copernicus DEM improves one arc second open global topography. IEEE Transactions on Geoscience & Remote Sensing. vol. 62, pp. 1-22, 2024, Art no. 4503922, https://doi.org/10.1109/TGRS.2024.3368015

Guth, P.L.; Trevisani, S.; Grohmann, C.H.; Lindsay, J.; Gesch, D.; Hawker, L.; Bielski, C. Ranking of 10 Global One-Arc-Second DEMs Reveals Limitations in Terrain Morphology Representation. Remote Sens. 2024, 16, 3273. https://doi.org/10.3390/rs16173273

Guth, P.L.; Van Niekerk, A.; Grohmann, C.H.; Muller, J.-P.; Hawker, L.; Florinsky, I.V.; Gesch, D.; Reuter, H.I.; Herrera-Cruz, V.; Riazanoff, S.; López-Vázquez, C.; Carabajal, C.C.; Albinet, C.; Strobl, P. Digital Elevation Models: Terminology and Definitions. Remote Sens. 2021, 13, 3581. https://doi.org/10.3390/rs13183581

Minár, J., Ian S. Evans, Marián Jenčo, 2020, A comprehensive system of definitions of land surface (topographic) curvatures, with implications for their application in geoscience modelling and prediction, Earth-Science Reviews, Volume 211, 103414, ISSN 0012-8252, https://doi.org/10.1016/j.earscirev.2020.103414

Strobl, P.A.; Bielski, C.; Guth, P.L.; Grohmann, C.H.; Muller, J.P.; López-Vázquez, C.; Gesch, D.B.; Amatulli, G.; Riazanoff, S.; Carabajal, C. The Digital Elevation Model Intercomparison eXperiment DEMIX, a community based approach at global DEM benchmarking. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2021, XLIII-B4-2021, 395–400. https://doi.org/10.5194/isprs-archives-XLIII-B4-2021-395-2021

Uhe, P., Lucas, C., Hawker, L., Brine, M., Wilkinson, H., Cooper, A., & Sampson, C. (2025). FathomDEM: an improved global terrain map using a hybrid vision transformer model. Environmental Research Letters, 20(3), 034002. https://doi.org/10.1088/1748-9326/ada972

These data contain bank strength measurements in a 100 km reach of the Solimões River from Sep/Oct 2022 recorded with a shear vane and a cohesion strength meter. In addition, processed satellite (Landsat) imagery from 1984-2021 was used to calculate erosional and depositional area in three 50-120 km long reaches in the Solimoes River presented here as shape-files. Processed Corona imagery 1967 for a 120 km long reach in the Solimoes River shows the banklines and bar outlines. A spreadsheet provides erosional and depositional area at 20-km sections along the 1,600km of the Solimoes River that were based on measuring floodplain width from a digital elevation model (FABDEM). We also attach a GeoTIFF file of the multibeam echo sounder (MBES) data collected during the field campaign in a 20-km long reach in the Solimoes River Full details about this dataset can be found at https://doi.org/10.5285/11786f86-a3ac-45ab-81b5-10fd157e3d7a

Deep learning-based super-resolution methods have been successfully applied to DEM downscaling studies by designing structures and loss functions of the model. However, the design of super-resolution models with enhanced performance remains a challenge, particularly in the context of hydrological studies, which require high-resolution DEMs with correct hydrological characteristics. In this study, we introduce a super-resolution model that integrates hydrologic knowledge (HKSRCGAN), aiming for the model to effectively maintain topographic features as well as the hydrologic availability of the DEMs. FABDEM with 30 m spatial resolution，which removes buildings and vegetation, was used in the experiment to demonstrate the usability of the proposed method. The hydrological knowledge derived from surface flow direction and hydrological features are integrated into a deep learning algorithm to guide model training. The results show that the HKSRCGAN outperforms the bicubic interpolation, SRCNN, SRGAN, SRResNet methods in reducing topographic errors and maintaining hydrologic characteristics. In the test area, the entropy difference analysis shows that the DEM generated by HKSRCGAN is more similar to the information contained in the reference DEM. Furthermore, super-resolution models integrating hydrological knowledge are valuable for modeling terrain shaped mainly by gravity and surface water flows. In addition, deep learning-based models integrating hydrologic knowledge are expected to be applied in DEM upscaling to maintain consistent hydrological characteristics.

FathomDEM Digital Elevation model for locations east of 30W (including Europe, Africa, Asia and part of Oceania).

FathomDEM is a global 30m DEM produced using a novel application of a hybrid vision transformer model. This model removes surface artifacts from a global radar DEM, Copernicus DEM, aligning it more closely with true topography.

This dataset is documented in the peer reviewed article FathomDEM: an improved global terrain map using a hybrid vision
transformer model

FathomDEM data for locations west of 30W is available in the dataset FathomDEM v1-0 Americas

The Global River Topology (GRIT) is a vector-based, global river network that not only represents the tributary components of the global drainage network but also the distributary ones, including multi-thread rivers, canals and delta distributaries. It is also the first global hydrography (excl. Antarctica and Greenland) produced at 30m raster resolution. It is created by merging Landsat-based river mask (GRWL) with elevation-generated streams to ensure a homogeneous drainage density outside of the river mask (rivers narrower than approx. 30m). Crucially, it uses a new 30m digital terrain model (FABDEM, based on TanDEM-X) that shows greater accuracy over the traditionally used SRTM derivatives. After vectorisation and pruning, directionality is assigned by a combination of elevation, flow angle, heuristic and continuity approaches (based on RivGraph). The network topology (lines and nodes, upstream/downstream IDs) is available as layers and attribute information in the GeoPackage files (readable by QGIS/ArcMap/GDAL).

A map of GRIT segments labelled with OSM river names is available here: Map with names

Report bugs and feedback

Your feedback and bug reports are welcome here: GRIT bug report form

The feedback may be used to improve and validate GRIT in future versions.

Regions

Vector files are provided in 7 regions with the following codes:

• AF - Africa
• AS - Asia (excl. Siberia)
• EU - Europe
• NA - North America
• SA - South America
• SI - Siberia
• SP - South Pacific/Australia

The domain polygons (GRITv06_domain_GLOBAL.gpkg.zip) provide 60 subcontinental catchment groups that are available as vector attributes. They allow for more fine-grained subsetting of data (e.g. with ogr2ogr --where and the domain attribute).

Vector files are provided both in the original equal-area Equal Earth Greenwich projection (EPSG:8857) as well as in geographic WGS84 coordinates (EPSG:4326).

Change log

• v0.6 - 2024-05-30
  • Rivers/streams outside of the GRWL mask forced by all OSM water lines (not only those with waterway=river/canal)
  • Some manual directions in the Irrawaddy delta and fixed erronous sink in the Volga delta
• v0.5 - 2024-02-14
  • Cyclicity and discontinuities resolved through improved algorithms, bug fixes, more sophisticated cycle solving algorithms and some manually forced directions. Only insignificant cycles (non-sinks, less than 50) were removed.
  • Added segment and reach attributes
  • Computational domain fixes
  • Segments include OSM river names
  • Asia domain split into Siberia and rest of Asia
  • Vector files available in EPSG:8857 and EPSG:4326
• v0.4 - 2023-03-11
  
  • First globally complete dataset published

Network segments

Lines between inlet, outlet, confluence and bifurcation nodes. Files have lines and nodes layers.

Attribute description of lines layer

Attribute description of nodes layer

Network reaches

Segment lines split to not exceed 1km in length, i.e. these lines will be shorter than 1km and longer than 500m unless the segment is shorter. A simplified version with no vertices between nodes is also provided. Files have lines and nodes layers.

Attribute description of lines layer

Accurate global digital elevation models (GDEMs) are essential for various geoscience applications. However, the accuracy of GDEMs in vegetated mountainous regions is relatively low due to substantial topographic relief and the penetration limitations of data acquisition techniques. The Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) acquires high-precision and high-density elevation measurements along its ground tracks on a global scale, offering reliable reference data for GDEM corrections. However, previous GDEM corrections using ICESat-2 altimetry data primarily focused on pixel-by-pixel vertical elevation corrections, often neglecting inter-pixel neighboring structure information, which is crucial for terrain modeling and analysis, particularly in areas with high relief. Therefore, this study incorporates not only ICESat-2 precise elevation observations and but also its along-track neighboring structure knowledge into a convolutional neural network (CNN)-transformer hybrid model, termed NSCT, to correct GDEMs in global scale. The 1 arc-second Copernicus DEM was selected as the target DEM for correction due to its demonstrated superior accuracy. The proposed NSCT model, trained on a diverse range of globally distributed areas with varying topography and vegetation, was evaluated using ICESat-2, global control points, and high-resolution DEMs. Its performance was compared against eight currently most used DEM correction models and publicly available corrected GDEM products, including FABDEM, FathomDEM, and GEDTM30. Correction results demonstrated that the NSCT model generally improved Copernicus DEM accuracy by 66.34%, and outperformed existing correction models in both vertical elevation and neighboring structure assessment across diverse topographic and vegetation conditions. Furthermore, validation using ICESat-2 data and high-resolution DEMs outside the training area, as well with its application to SRTM and AW3D30 GDEM, demonstrated that the NSCT model exhibited superior transferability capability and consistently outstanding performance. This study is the first to integrate the precise elevation observations of ICESat-2 with along-track neighboring structure knowledge into GDEM corrections, offering valuable insights for future research in terrain modeling and analysis.

The Qinghai Tibet Plateau is an important ecological security barrier area. Clarifying the suitable space for construction land on the Qinghai Tibet Plateau is of great significance for coordinating the relationship between ecological protection and human activities, and promoting the optimization of spatial layout of population and industry. Based on the land resource suitability evaluation model for human activities and the bare land digital elevation model FABDEM for removing forests and buildings, a 30m spatial resolution construction land suitability spatial distribution dataset covering the entire Qinghai Tibet Plateau was produced. By setting grading thresholds of 0.73, 0.58, 0.38, and 0.26, the suitable, relatively suitable, generally suitable, unsuitable, and unsuitable areas for construction land on the Qinghai Tibet Plateau were determined to be 3593 km ², 100258 km ², 166033 km ², 208411 km ², and 2104249 km ², respectively, accounting for 0.14%, 3.88%, 6.43%, 8.07%, and 81.48% of the total land area. The accuracy of the actual construction map layer was verified using the 2020 China Multi period Land Use Remote Sensing Monitoring Data (CNLUCC) and GlobeLand30 surface cover data. The dataset has high explanatory power for the spatial distribution of actual construction land, with overall accuracies of 76.41% and 81.65%, respectively. The dataset consists of two data layers: the suitability index of construction land and the suitability level of construction land, stored in FGDBR raster data format with a spatial resolution of 30 m. The data type of the suitability index of construction land is a 32-bit unsigned integer, with data validity ranging from 0 to 782200 and a scaling factor of 0.000001. The larger the index, the higher the suitability, and the smaller the index, the lower the suitability; The suitability level data layer for construction land is a 4-digit unsigned integer. Based on the grading thresholds of 0.73, 0.58, 0.38, and 0.26, the suitability index is classified into suitable, relatively suitable, generally suitable, unsuitable, and unsuitable levels, represented by integers from 1 to 5. This dataset can provide decision-making references for the relocation of ecological immigrants on the Qinghai Tibet Plateau, the construction of stable and solid border towns and villages, and the optimization of national spatial development and protection patterns. The dataset was originally published in the Science Data Bank.