CoastalDEM v2.1: A high-accuracy and high-resolution global coastal elevation model trained on ICESat-2 satellite lidar

CoastalDEM v2.1 is a high-resolution global digital elevation model (DEM) designed to improve the accuracy of coastal flood and sea level rise assessments. By employing a convolutional neural network (CNN) trained on NASA's ICESat-2 satellite lidar data and using NASADEM as a base, it significantly reduces the positive vertical bias common in satellite radar DEMs, particularly in densely populated urban areas where buildings often distort elevation readings. Validation against ICESat-2 and airborne lidar data from the US and Australia indicates that CoastalDEM v2.1 outperforms other leading global models like TanDEM-X and MERIT, especially in the critical elevation band below 5 meters.

• CoastalDEM v2.1 demonstrates superior global accuracy compared to other publicly available global-scale models, particularly in low-lying and densely populated areas.
• In the elevation band below 5 meters across all areas, the 30 m version of CoastalDEM v2.1 reduces global median bias to less than 0.01 m, with a root mean square error (RMSE) of 2.63 m and a 90th percentile linear error (LE90) of 2.99 m.
• For coastal areas with moderate development (population density >1,000 people per km2) and elevations below 5 m, mean vertical bias improved from -0.5 m in version 1.1 to -0.1 m in version 2.1, representing a 91-95% bias reduction compared to other comparable DEMs.
• CoastalDEM v2.1 shows consistent performance across different nations, with a lower median bias in 90% of countries and lower RMSE in at least 78% of countries when evaluating areas with population density >1,000 people per km2 and elevation <5 m.
• Validation using airborne lidar data in the US and Australia confirms that CoastalDEM v2.1 has a median bias closer to zero and lower RMSE/LE90 values in the <5 m elevation band than competing global DEMs.
• The model specifically improved accuracy in south Florida, particularly around Miami, where previous global DEMs exhibited multi-meter biases due to dense development and vegetation.

APA

Kulp, S. A., & Strauss, B. H. (2021). CoastalDEM v2.1: A high-accuracy and high-resolution global coastal elevation model trained on ICESat-2 satellite lidar. Climate Central. https://assets.ctfassets.net/cxgxgstp8r5d/3f1LzJSnp7ZjFD4loDYnrA/71eaba2b8f8d642dd9a7e6581dce0c66/CoastalDEM_2.1_Scientific_Report_.pdf

Chicago

Kulp, Scott A., and Benjamin H. Strauss. CoastalDEM v2.1: A high-accuracy and high-resolution global coastal elevation model trained on ICESat-2 satellite lidar. Climate Central, 2021. https://assets.ctfassets.net/cxgxgstp8r5d/3f1LzJSnp7ZjFD4loDYnrA/71eaba2b8f8d642dd9a7e6581dce0c66/CoastalDEM_2.1_Scientific_Report_.pdf.

Wikipedia

{{cite report |last1=Kulp |first1=Scott A. |last2=Strauss |first2=Benjamin H. |title=CoastalDEM v2.1: A high-accuracy and high-resolution global coastal elevation model trained on ICESat-2 satellite lidar |publisher=Climate Central |date=21 September 2021 |url=https://assets.ctfassets.net/cxgxgstp8r5d/3f1LzJSnp7ZjFD4loDYnrA/71eaba2b8f8d642dd9a7e6581dce0c66/CoastalDEM_2.1_Scientific_Report_.pdf |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{kulp2021coastaldem, author = {Kulp, Scott A. and Strauss, Benjamin H.}, title = {{CoastalDEM v2.1: A high-accuracy and high-resolution global coastal elevation model trained on ICESat-2 satellite lidar}}, institution = {Climate Central}, year = {2021}, month = sep, url = {https://assets.ctfassets.net/cxgxgstp8r5d/3f1LzJSnp7ZjFD4loDYnrA/71eaba2b8f8d642dd9a7e6581dce0c66/CoastalDEM_2.1_Scientific_Report_.pdf}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }