Regional “Bare-Earth” Digital Terrain Model for Costa Rica Based on NASADEM Corrected for Vegetation Bias

Pimenova, Olga; Roberts, Craig; Rizos, Chris

doi:10.3390/rs14102421

Cited by 3 publications

(1 citation statement)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, efforts have been underway to generate a true bare-earth model from existing open-access global DEMs by using lidar point clouds and machine learning methods [7][8][9][10][11][12][13]. Multi-Error-Removed Improved Terrain (MERIT) DEM was generated by applying a global-scale error-removal algorithm using the existing space-borne DEMs like SRTM and ALOS World 3D DEM (AW3D DEM) by Yamazaki et al [14].…”

Section: Introductionmentioning

confidence: 99%

Validation of MERIT DEM’s Performance as a Bare-Earth Model Using ICESat-2 Geolocated Photons

Dandabathula,

Hari,

Sharma

et al. 2023

EARTH

View full text Add to dashboard Cite

Digital elevation models represent the Earth's surface and play a key role in earth sciences by enabling the possibility of deriving terrain variables; the terrain variables are essential inputs for environmental modeling. The availability of open-access digital surface models has significantly advanced the understanding of earth system dynamics and also allowed researchers to generate digital terrain models, aka bare-earth models. These bare-earth models are essential data sets for applications related to hydrology and geomorphology, especially for disaster management. Under the category of open-accessible bare-earth models, Multi-Error-Removed Improved-Terrain DEM or MERIT DEM is the first kind of product unfolded by applying numerous error removal algorithms from existing DEM sources. This research reports the results after validating the MERIT DEM's performance by emphasizing its tree-height bias removal algorithm. Towards this, ground-reflected photons accrued from the ICESat-2 mission were used as reference data due to their attribution of high accuracy. Two test sites, one located in the rugged terrain of the outer Himalayas, the Lacchiwala Reserve forest, and the other, rolling hills at the Bhadra wildlife sanctuary located in the Western Ghats of the Indian sub-continent were used as test sites for validating the MERIT DEM's accuracy. The results derived after computing statistical formulae like RMSE, MAE, MBE, and profile-based visual analytics helped understand the performance of the MERIT DEM as a bare-earth model. The RMSE, MAE, and MBE for the Lachhiwala Reserve forest are 10.28 m, 7.78 m, and 0.69 m, respectively. Similarly, the RMSE, MAE, and MBE values for the Bhadra wildlife sanctuary are 4.52 m, 3.82 m, and 3.04 m, respectively. The assessment confirms that the accuracies are within the MERIT DEM's specifications and assured the successful implementation of MERIT DEM's tree-height removal algorithm since the elevations from the MERIT DEM are always lesser than the canopy height in both the test sites. Our research also investigated the reasons for the inaccuracies obtained at both the test sites and suggested using improved tree-height estimations from high-resolution canopy height data in the future version of MERIT DEM.

show abstract