Digital elevation model quality on digital soil mapping prediction accuracy

Abstract

Digital elevation models (DEM) used in digital soil mapping (DSM) are commonly selected based on measures and indicators (quality criteria) that are thought to reflect how well a given DEM represents the terrain surface. The hypothesis is that the more accurate a DEM, the more accurate will be the DSM predictions. The objective of this study was to assess different criteria to identify the DEM that delivers the most accurate DSM predictions. A set of 10 criteria were used to evaluate the quality of nine DEMs constructed with different data sources, processing routines and three resolutions (5, 20, and 30 m). Multinomial logistic regression models were calibrated using 157 soil observations and terrain attributes derived from each DEM. Soil class predictions were validated using leave-one-out cross-validation. Results showed that, for each resolution, the quality criteria are useful to identify the DEM that more accurately represents the terrain surface. However, for all three resolutions, the most accurate DEM did not produce the most accurate DSM predictions. With the 20-m resolution DEMs, DSM predictions were five percentage points less accurate when using the more accurate DEM. The 5-m resolution was the most accurate DEM overall and resulted in DSM predictions with 44% accuracy; this value was equal to that obtained with two coarser resolution, lower accuracy DEMs. Thus, identifying the truly best DEM for DSM requires assessment of the accuracy of DSM predictions using some form of external validation, because not necessarily the most accurate DEM will produce the best DSM predictions.

Optimal resolution of soil properties maps varies according to their geographical extent and location

“…They show a significant effect of cell size on predictions (Usery et al, 2004;Wu et al, 2005), and globally an increase of performance with decreasing resolution (Guo et al, 2019;Lassueur et al, 2006;Zhang and Montgomery, 1994). Other studies also suggest no effect of resolution, or better performance of coarser resolution (Costa et al, 2018;Kim et al, 2014;Roecker and Thompson, 2010). An effect of pixel aggregation from small size to coarse resolution showed both improved and decreased digital map accuracy (Carmel, 2005;McCloy and Bocher, 2007).…”

Digital assessments of soil organic carbon storage using digital maps provided by static and dynamic environmental covariates

“…In this study, we derived 17 terrain attributes from a DEM at a resolution of 30 m using SAGA GIS (Conrad et al, 2015). The quality of DEM influences the accuracy of terrain attributes (Costa et al, 2018). The SRTM (The Shuttle Radar Topography Mission) data (USGS, 2018) was used as DEM source.…”

Digital soil class mapping in Brazil: a systematic review

“…Of the 334 studies, 263 presented cartographic scale and spatial resolution (pixel size) information used (Table 3); 38 studies were found incompatible with the planimetric PEC-PCD, since their pixel size is higher than that indicated at the cartographic scale. (Giasson et al, 2006) 2008 (Figueiredo et al, 2008) 2009 (Crivelenti et al, 2009) 2010 (Chagas et al, 2010); (Coelho and Giasson, 2010) (Höfig et al, 2014); (Teske et al, 2014) 2015 (Bagatini et al, 2015); (Giasson et al, 2015); (Teske et al, 2015a); (Teske et al, 2015b); (Vasques et al, 2015) 2016 (Arruda et al, 2016); (Bagatini et al, 2016); (Demattê et al, 2016); (Dias et al, 2016); (Henrique et al, 2016); (Pelegrino et al, 2016) 2017 (Chagas et al, 2017); (Wolski et al, 2017) 2018 (Costa et al, 2018); (Meier et al, 2018) 2019 (Campos et al, 2019a); (Campos et al, 2019b); (Moura-Bueno et al, 2019); (Silva et al, 2019); (Silvero et al, 2019) All learning algorithms were assigned to a type of classifier such as Artificial Neural Network (ANN), Bayes classifiers, Decision Tree (DT), Logistic Regression (LR) and Support Vector Machine (SVM). Approximately 95 % of the studies used DT, ANN and LR classifiers (Table 4).…”