Literature research and evaluation
We conducted a systematic and reproducible literature research and evaluation, adhering to standards of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) process. All details of the literature research, selection and evaluation are provided in Supplementary Tables 1 and 2. We used the Scopus search engine and used a search term for titles, abstracts and keywords that covered keywords from low-lying coastal environments, processes and hazards, as well as elevation information and elevation dynamics, impact and/or hazard assessments, projections or evaluations (Supplementary Table 1). The Scopus search engine was used as it constitutes one of the most transparent search engines. We are aware of the limitation that Scopus does not include the entire available (scientific) literature and therefore might introduce a bias into our investigation. Therefore, we performed a second independent literature evaluation on 96 additional studies (90 after excluding publications by the authors; non-systematically collected through other search engines) (Extended Data Fig. 8) to ensure the results from the search were bias-free. All results presented in the main paper are based on the Scopus-based systematically researched literature only. From the initial 35,001 documents that were identified through Scopus, we included only peer-reviewed scientific articles, reviews including original analyses, or data papers. Moreover, we limited our search to the most recent 15 years (2009–2025) to include only the most recent research from the period during which both global DEMs and global MDT data were available and well-established, thereby ensuring a levelled playing field of global data availability across the globe to evaluate datum conversion and sea-level reference methodologies for assessments. Further filters applied refer to the exclusion of certain unrelated subject areas, keywords and limiting the search to the English language. This refined search term reduced the amount of literature to 7,241 documents that were subsequently screened in two consecutive rounds following established inclusion and exclusion criteria (Supplementary Fig. 1 and Supplementary Table 1). The first screening was conducted based on title, abstract and keywords, and the second screening was conducted based on in-depth analysis of the full article and, when needed, its references and supplementary material. The screening was consistently performed by K.S., whereas a subset of 1,000 articles (first screening) was cross-checked and validated by P.S.J.M., hereby not encountering conflicting decisions. Publications that were categorized as ‘yes’ and ‘maybe’ in the first screening were included in the second, full-article screening. The inclusion criteria were focus on coastal land and coastal sea level, focus on coastal hazard and impact assessment, focus on coastal lowlands and the use of satellite-borne elevation information. The main exclusion criteria were the absence of elevation data use, use of airborne lidar without combination with satellite-borne elevation data, methodological and technical articles without application, articles referring to elevation data cited in tertiary source or beyond (only primary or secondary sourced studies included) and focus on secondary hazards (cascading effects). The finally included literature, consisting of 385 publications, underwent an in-depth evaluation of their respective research data and methodology, scrutinizing them, and—where needed—also their secondary references, to assess all relevant details related to study area, distance to shore, study focus and/or flood type, number of DEMs, DEMs used, actuality of DEM, DEM type, DEM coverage, acquisition and technique, performance of DEM accuracy assessment, DEM accuracy, spatial resolution, DEM availability, horizontal datum used, vertical datum documentation, vertical datum conversion, vertical datum type, vertical datum used, implementation of sea-level reference, and—where needed—additional comments (Supplementary Table 2).
The individual papers were accessed by either the Scopus search engine, ISI Web of Science, Science Direct, Google Scholar or ResearchGate. The complete evaluation of the selected 385 publications and the subsequent analysis of the results following the developed evaluation criteria and coding are provided in Supplementary Table 2. Only publications that handle all datasets in a consistent vertical reference frame (that is, performing a correct vertical datum conversion) and refer to the latest available sea-level information at their time (that is, using an up-to-date sea-level reference and correct vertical datum conversion) were considered as proper. As vertical datum conversion always requires additional data next to a DEM-based coastal elevation dataset (for example, vertical datum information, offset between geoid/geoids and tidal level, sea-level height data) as well as a conversion procedure (including a specific GIS or coding environment or conversion service), we expect researchers who perform a datum conversion to document the additional datasets and procedures in their papers and/or supplementary information or supplementary data. In case a paper does not provide any documentation of additional datasets or required conversion steps, we presume that no datum conversion was performed. This assumption was confirmed to be correct for numerous papers for which the absence of datum conversion becomes apparent in the results of the study.
Finally, we evaluated more than 90 additional publications that fitted the selection criteria, but were not included in the Scopus-searched literature. This subset was separately collected by the authors through additional, non-systematic literature searches, for example, to include relevant IPCC-referenced assessments, and served as an independent dataset to evaluate potential bias present in our systematic Scopus-searched literature dataset (Supplementary Data 1). The additional literature dataset underwent the same scrutinizing evaluation (Supplementary Table 1). We found nearly equivalent findings on the occurrence and percentage distribution for the various vertical datum and sea-level reference issues (Extended Data Fig. 8) as for the systematic review (Fig. 1), which suggests that our Scopus-based systematic review is unbiased and provides representative results for the existing body of literature.
Vertical datum conversion and processing of DEMs
To obtain land elevation above continuous local sea level globally and quantify the discrepancies to land elevation with respect to global geoids or datum conversion errors, we processed four of the most recent DEMs at a global scale by converting them from their original vertical reference system to MSL as indicated by the MDT product. The DEMs included were CoastalDEM v.2.1 (original vertical reference system: EGM96, spatial resolution: 90 m × 90 m)40, FABDEM v.1.0 (original vertical reference system: EGM2008, spatial resolution: 30 m × 30 m)41, GLL-DTM v.2 (original vertical reference system: MDT53, spatial resolution: 1° × 1°, that is, about 1,000 m × 1,000 m)42 and DeltaDTM v.1 (original vertical reference system: EGM2008, spatial resolution: 30 m × 30 m)43. The publicly available GLL-DTM v.2 (referenced to CNES-CLS13 MDT53) contains a geoid conversion error (that is, the EGM-DIR R4 geoid was assumed to equal the EGM96) and does not apply the latest available MDT dataset, at the time. Therefore, we obtained the pre-converted GLL-DTM v.2 (referenced to EGM96) from the authors and conducted the vertical datum conversion to MSL44.
We used the latest available global MDT HYBRID-CNES-CLS2022 dataset44, which provides sea surface height above geoid (GOCO06s) across the globe, measured by satellite altimetry and combined with gravitational field information, oceanographic data from drifting buoys, high-frequency radar velocities and hydrological profiles. The MDT dataset44 provides spatially continuous information of sea surface height above the GOCO06s geoid at a resolution of 0.125° averaged over a period from 1993 to 2021. Therefore, it provides the latest available information on global MSL and an accurate substitute for local tide gauge information in those regions. Several studies confirm the accuracy of MDT data in the range of cm (see, for example, refs. 54,55), although vertical accuracy decreases up to about 4 cm within 10 km of the coast56. It serves as an open-accessible product that—if properly aligned with elevation information—can be used to adjust elevation with respect to local sea level continuously along the coastlines of the world.
The following section documents a proper, consistent and reproducible vertical datum conversion of the four DEMs, providing all required datasets, datum information, processing steps and used software environments and may serve as an example for future studies or existing studies aiming to re-evaluate previous assessments. For all our computations, we used the ArcGIS Pro environment. To reference coastal elevation of the DEMs to MSL as given by MDT, the offsets of the underlying respective vertical reference systems (that is, EGM96, EGM2008 and GOCO06s) were determined (Extended Data Fig. 2).
Geoid information was obtained from the openly accessible calculation service of the International Centre for Global Earth Models from GFZ Helmholtz Centre for Geosciences57. Point data on geoid height anomaly to the WGS84 ellipsoid was obtained for the entire globe at a resolution of 0.085° for the EGM96, EGM2008 and GOCO06s geoids, respectively. For each geoid, the height anomaly points were interpolated into a global raster by using multiquadric radial basis functions, which gave the most accurate interpolation results and is also used by gravitational field modelling studies58,59. Subsequently, the global geoid raster files were resampled to a common spatial resolution of 90 m × 90 m (for CoastalDEM v.2.1, FABDEM v.1.0 and DeltaDTM v.1) and 1,000 m × 1,000 m (for GLL-DTM v2) by using bilinear resampling to be comparable with the spatial resolution of the DEMs throughout the entire datum conversion process. Geoid offsets were determined by subtracting the GOCO06s geoid height anomaly raster from the EGM96 and EGM2008 geoid height anomaly rasters.
Using MSL as indicated by MDT as a vertical datum for land elevation data requires the extrapolation of MDT data over land. We extracted point values from the MDT HYBRID-CNES-CLS2022 raster using bilinear interpolation at point locations and subsequently extrapolated them over land using an inverse distance weighting algorithm and a smooth neighbourhood type with a smoothing factor of 0.5. The resulting raster dataset was resampled to two spatial resolutions of 90 m × 90 m (for CoastalDEM v.2.1, FABDEM v.1.0 and DeltaDTM v.1) and 1,000 m × 1,000 m (for GLL-DTM v.2) by using bilinear resampling to enable comparability with the spatial resolution of the DEMs and to avoid the introduction of potential artefacts stemming from large differences in spatial resolution in the datum conversion process.
... continue reading