Fig 1.
Study plot with sampling scheme.
A. The study plot (77 ha) is a part of a larger agricultural plot (176 ha), it is located in the Northern part of the Czech Republic. Soil survey with regular soil sampling scheme was preceded and the sampling points were randomly divided to training and validation subsets. B. Geomorphologic features of the study plot (vertical scale is 2.5x to horizontal scale). The plot is formed by two perpendicular side valleys and a plateau. The transition between these two units is characterized by rather distinct convex shoulders.
Fig 2.
Typical soil toposequence of the study plot.
Luvisols (LU) are present on flat or slightly sloping areas (plateau). Regosols (RG) result from accelerated erosion on the steeper convex parts of the landscape. The removed material is transported down the slope and sediments in the base of the side valleys. If the depth of the colluvial material is higher than 50 cm, the soils are classified as Colluvial soils (CO), if less as Colluvic Luvisols (LUco).
Table 1.
Basic statistic of the terrain derivatives.
The table presents mean, median, min., max and standard deviation values of the six terrain derivatives for all resolutions (1, 2, 3, 5, 10, 20 and 30 meters).
Fig 3.
Terrain derivatives means with 95% confidence interval for the different raster resolutions: altitude (ALT), slope (SLP), plane curvature (PLANC), topographic position index (TPI), LS factor (LS) and convergence index (CONVIN).
Table 2.
Statistically significant differences between the terrain derivatives means (Post hoc LSD test).
Fig 4.
Soil units predicted by the four models at given resolutions 1, 2, 3, 5, 10, 20 and 30 meters.
(CO-blue; LU-green; RG-red colour).
Fig 5.
Models accuracy at different raster resolution for all soil classes (left column) and Colluvial soils only (right column).
The dark grey vertical line at each column indicates the range (min-max) of all four models at a given resolution. The short horizontal line is the mean value of the four models accuracy.
Fig 6.
Proportion of soil units on the soil cover of the study plot calculated by the four models at all raster resolutions.
The blue and red lines indicate the spatial extent of CO and RG respectively according the soil survey (nr. of observed profiles).
Fig 7.
Deviation of CO soils prediction by models from an aerial extent of CO found by the soil survey (relative %).
Blue columns show negative and positive relative % deviation. The light blue columns indicate absolute average deviation.