Global or quasi-global open-access digital elevation models (DEMs) produced from satellite data, are marked by continuous improvements in quality and accuracy (Gesch, 2018; Kakavas, Kyriou, & Nikolakopoulos, 2020; Mudd, 2020; Guth et al., 2021). The most common and popular global DEMs include versions of SRTM DEM (Farr et al., 2007), ASTER GDEM (Abrams, Crippen, & Fujisada, 2020), ALOS AW3D30 DSM (Tadono et al., 2016), MERIT DEM (Yamazaki et al., 2017), and Copernicus GLO-30 DEM (Airbus, 2020; Strobl, 2020)[1].
Global DEMs, generally more similar to digital surface models (DSMs) than to digital terrain models (DTMs), are frequently used in geoenvironmental and geoengineering studies. They have been adopted for multiple research purposes including the analysis and modeling of geoenvironmental processes related to terrain morphology more than to the overall surface morphology, which also includes anthropic structures and vegetation (Guth et al., 2021). Terrain morphology is fundamental in the context of geostructural, geomorphological, and geoenvironmental interpretation. In many applications, morphometric variables derived from DEMs are used as input features in supervised learning approaches or, at least, as proxies of other environmental variables (Wilson & Gallant, 2000; Hengl & Reuter, 2009; Florinsky, 2016, 2021; Minár, Krcho, & Evans, 2016; Trevisani & Florinsky, 2021; Trevisani et al., 2021). Examples of DEM-based studies include, but are not limited to, landslides susceptibility models (Meena & Nachappa, 2019; Titti et al., 2021), derivation of shear wave seismic velocity of shallow subsoil (Wald & Allen, 2007; Thompson, Wald, & Worden, 2014; Heath et al., 2020), derivation of drainage networks and basins with a possible adoption of ad-hoc conditioning procedures of the drainage network in anthropized areas (Cavalli et al., 2013). Another area of DEM-based research concerns geodiversity studies, where morphometric parameters contribute to derive indexes of geodiversity (Melelli, 2014; Chrobak, Novotný, & Struś, 2021). Frequently, geodiversity studies covering large areas use global DEMs, despite the focus is on terrain morphology. DEMs are also fundamental for many processes concerning remote sensing methodologies, for example, in the context of SAR interferometry technology in the study of ground displacements (Bayer, Schmidt, & Simoni, 2017).
Given the importance of global DEMs (Schumann & Bates, 2018), it is not surprising that the wide set of studies concerning the quality and accuracy of these digital products (Purinton & Bookhagen, 2017; Rizzoli et al., 2017; Boulton and Stokes, 2018; Caglar et al., 2018; Florinsky, Skrypitsyna, & Luschikova, 2018; Gonzalez & Rizzoli, 2018; Grohmann, 2018; Jain et al., 2018; Li & Zhao, 2018; Florinsky et al., 2019; Liu et al., 2019; González-Moradas & Viveen, 2020; Kakavas, Kyriou, & Nikolakopoulos, 2020; Vassilaki & Stamos, 2020; Mesa-Mingorance & Ariza-López, 2020; Guth & Geoffroy, 2021), including the recent DEMIX initiative (Strobl et al., 2021) aimed to develop a participative and standardized approach for global DEMs quality evaluation. Most of the studies generally focus on the evaluation of the inherent quality of DEMs products considering a reference DEM or a set of ground control points (Guth et al., 2021). However, as noted above, for geoengineering and geoenvironmental applications, it is fundamental to understand the capability and quality of global DEMs from the viewpoint of their capability to describe terrain morphology, possibly considering different land covers and morphological settings. In fact, in areas with dense vegetation and anthropic structures, the capability of global DEMs to reproduce terrain morphology is mainly influenced by the land cover characteristics, while in other areas it is mainly affected by the characteristics of local morphology, such as slope and surface roughness.
The capability of a DEM to reproduce terrain morphology cannot be fully captured by the statistical analysis of pixel-by-pixel differences respect to a reference DEM, adopting conventional metrics such as root mean square error (RMSE), mean/median error and mean/median absolute error (e.g., Florinsky et al., 2019; Crema et al. 2020; Polidori & El Hage, 2020). Differently, the spatial variability structure (Isaaks & Srivastava, 1989) of DEMs should be considered in the evaluation of their quality, for example including the capability to reproduce fine-scale morphology and different aspects of surface roughness (e.g., Trevisani, Cavalli, & Marchi, 2012). The necessity to move beyond the evaluation of error statistics on a pixel-by-pixels basis is advocated also in the analogous context of image analysis. From this perspective, the structural similarity index (Wang et al., 2004; Crema et al., 2020) could be mentioned, since it considers the spatial variability structure in evaluating the similarity between a sample image and a reference one.
Accordingly, this research focuses on the quality evaluation of two of the best 1 arc-second public global digital elevation models, Copernicus GLO-30 version 2021_1 (hereinafter referred to as COP) (Airbus, 2020; Strobl, 2020) and ALOS AW3D30 version 3.2 (hereinafter referred to as ALOS) (JAXA, 2021a, 2021b), from the perspective of their capability to represent the terrain morphology of a complex alpine landscape, located in the alpine Trentino Province (Italian Alps). The analysis, following a previous study (Florinsky et al., 2019) concerning the quality evaluation of several global DEMs in the same area, is conducted considering a reference DEM (herewith referred to as reference) derived from the upscaling of an accurate and high-resolution airborne Lidar DEM (Cavalli et al., 2013; Portale cartografico, 2016). The study area is well suited for an analysis of this kind for several reasons: 1) large extent, approximately 6210 km2; 2) high heterogeneity in morphology and land cover; 3) demanding morphological conditions for the remote sensing technologies adopted for the derivation of the global DEMs; 4) public availability of reference DEMs with creative common license (Portale cartografico, 2016).
The two global DEMs are considered among the best freely available global DEMs at 1 arc-second resolution. On the one hand, COP version 2021_1 seems particularly promising (Guth & Geoffroy, 2021) with remarkable improvements respect to the older EU-DEM version 1.1 (Bashfield & Keim, 2011). On the other hand, ALOS resulted the one with the best performances in the previous study conducted in the Alpine area (Florinsky et al., 2019) as well as in other regions (Caglar et al., 2018; Florinky, Skrypitsyna, & Luschikova, 2018; Jain et al., 2018; Li & Zhao, 2018; Liu et al., 2019; González-Moradas & Viveen, 2020). The quantitative analysis of the DEMs quality is performed both in terms of conventional error statistics computed on a pixel-by-pixel basis as well as from the perspective of the capability to represent the spatial variability structure (i.e., roughness) of the local surface morphology. Moreover, the analysis is performed considering specific land cover types and morphological settings, permitting to evaluate the capability of the two global DEMs to reproduce terrain morphology in different land cover and morphological settings.
[1] In this article, we use the term DEM according to its definition given by Guth et al. (2021): ‘DEM (digital elevation model): General term for a digital representation of elevations (or height) of a topographic surface in form of a georectified point-based or area-based grid, covering the Earth or other solid celestial bodies’. According to other definitions reported in Guth et al. (2021), the listed global DEMs are more similar to digital surface models (DSMs), that is, DEMs comprising buildings and vegetation, rather than to digital terrain models (DTMs), that is, DEMs with buildings and vegetation cover removed (‘bare-earth’ DEMs).