On Copernicus DEM - DTM and ICESat for height and biomass retrieval of tropical forests

Journal article, 2026

The Copernicus DEM - DTM difference was shown by Soja et al. in [1] to be linearly correlated with above ground biomass, AGB, for eleven test sites with a resolution of about 30 m for the Copernicus DEM and 100 m for the DTM. The goal of this paper is to give a model to explain why this is caused by both variables being dependent on the product of forest density and height. An expression representing the mean of TanDEM-X phase heights or Copernicus DEM - DTM according to the Interferometric Water Cloud Model (IWCM) is used to estimate canopy height, area-fill (canopy cover), and AGB, the latter with and without 5% training data. The IWCM area-fill is determined using an ICESat-type approach [2], introducing a parameter q0, determined by the geolocation of the site, to quantify the heightdensity relationship. The analysis is conducted based on the data from [1] using LVIS lidar observations from the TropiSAR and AfriSAR sites and ALS data from the BIOSAR sites (Krycklan and Remningstorp) and various methods from three additional sites, altogether from four continents. Results indicate an almost linear relationship, between mean phase height, IWCM mean canopy height, and AGB. For the boreal sites, the relative RMSE is comparatively high, whereas for the tropical sites the relative RMSE is around 14% for mean canopy height and 11-18 % for the tropical sites but for Mondah (31%) for AGB in line with [1]. The lower RMSEr for sites with larger forest mean heights is mainly associated with a relatively constant Copernicus DEM vertical error. The almost linear relation between mean values of the Copernicus DEM - DTM difference and the mean AGB with r²=0.97 demonstrates the feasibility of AGB retrieval without local calibration, tested on three temperate sites on two continents. All results are based on the availability of a DTM, which is expected from BIOMASS.