Selection of optimal scales for soil depth prediction on headwater hillslopes: A modeling approach

摘要: Abstract Landform attributes derived from digital elevation models (DEMs) are the most commonly used factors to predict soil depths on hillslopes. However, selection of an appropriate algorithm calculate terrain and identify optimal DEM resolution remains ambiguous. In this study, we propose a method use high-resolution spatial depth data obtain at hillslope scales. A geophysical (ground penetrating radar, GPR) was investigate reasons for findings. Point-scale 116 sites were collected two adjacent headwater hillslopes (H1: 0.42 ha H2: 0.31 ha) Hemuqiao hydrological experimental station in Southeast China. The datasets using total variable spacing level then DEMs nine resolutions: 0.25, 0.50, 0.75, 1.00, 2.00, 3.50, 5.00, 7.50 10.00 m. Nine primary secondary topographic derived. Two different algorithms (D8 D ∞) calculating contributing areas related compared. We compared both linear (multiple regression, MLR) non-linear (artificial neural network, ANN) prediction. Results demonstrated that performed well predicting depth. Specifically, MLR better than model ANNs. Additionally, found multiple-direction (D ∞) allowed flow divergence avoided abrupt changes predictions (orphan cells) should be adopted construction. D ∞ divergent areas, such as ridges side slopes, also worked convergent valleys. Moreover, our results moderate (e.g., 2.00 m) attributes, instead finest resolution, achieved best prediction with lowest root mean square error (RMSE) absolute (MAE) highest values coefficient determination (R2). GPR indicated valley accumulated more soils side-slope sharp increase valley. Comparing width obtained by GPR, average (AVW) considered good measure choosing