SPATIAL DISTRIBUTION OF SOIL DEPTH IN A HEADWATER CATCHMENT UNDERLAIN BY WEATHERED GRANITE

Soil depth is a crucial input parameter for the modeling of slope stability. Although a penetration test using a cone penetrometer is one of the standard techniques to measure soil depths, this method is both labor and time consuming. Therefore, some previous studies have attempted to estimate soil depths using topographic information such as inclination angle and catchment area. However, the topographic technique has inadequate prediction accuracy in most headwater catchments. In this study, we conducted precise measurements of soil depth distributions in a headwater catchment and evaluated the applicability of kriging methods.

We selected a small catchment (2,349 m²) underlain by weathered granite bedrock material in Mt. Rokkou, Japan. First, we measured the soil depth using a cone penetrometer at 36 points, in grids with a spacing of 10 m. We then measured the soil depth at an additional 40 points located between the original grid points. To apply kriging methods, we chose 70, 60, 50, 36, 25, and 19 points, respectively, from the total of 76 observation points. On the basis of the observed soil depths at the chosen points, we predicted soil depths for the remaining points. The prediction accuracy was evaluated by calculating a coefficient of determination.

The soil depth distribution, derived from the measurements at the 36 grid points and the additional 40 points, revealed two anomalous thick soil depths, one on the left side of the valley (2.4 m) and the other on the right side (2.14 m). In the soil depth distribution derived from only the measurements at the 36 grid points, the anomalous soil depth in the right valley disappeared. Moreover, the measurements at the 36 grid points could not indicate differences in soil depth distributions between the left and right banks. Soils were very shallow (about 0.2 m) on the left bank but thick on the right bank (about 2.3 m). We concluded that measurements using grids with a spacing of 10 m were not dense enough to fully evaluate the soil depth distribution in this catchment. Kriging estimates based on the 19 and 25 observation points produced poor results. When the number of the observation points was increased to 70, the kriging estimates produced a relatively high value (0.5) of the coefficient of determination.