Using Digital Elevation Models as an Environmental Predictor for Soil Clay Contents

The objective of this study was to evaluate the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) as an environmental predictor for soil clay content (SCC). It was based on the applicability of different DEMs, i.e., SRTM with 90-m resolution and airborne Light Detection and Ranging (LIDAR) (in 24- and 90-m resolution), using regression-tree analysis. Ten terrain parameters were generated from these DEMs. These terrain parameters were used along other environmental variables to statistically explain SCC content in Denmark. Results indicated that the SRTM tree model (T1: 90-m resolution) explained the variability of SCC measurements quasi-similarly (variance V = 60%) to the LIDAR tree models with 24-m (T2) or 90-m (T3) resolution (V = 60% for 12 and 61.5% for T3). The prediction performances (in terms of RMSE) of the produced maps (using these trees) compared with independent field observations from the validation data set (9000 sites) were estimated as follows: Map T1, RMSE = 3.57%; Map T2, RMSE = 3.25%; and Map T3, RMSE = 3.15%. The relative improvement of T2 compared with Ti or 13 varied between 8.96 and 11.76%, respectively. Independent validation data also reflected higher correlations between measured SCC and SCC predicted from 12 (R-2 = 0.60) compared with the other tree models (T1, R-2 = 0.56; T3, R-2 = 0.54). The modeling results indicate that the SRTM (including derivatives) has less predictive power than the LIDAR DEMs (with different resolutions) for mapping SCC in a low-relief landscape in Denmark.