Report No. : GEO Report No. 144
Report Title : Correlations between Rainfall, Landslide Frequency and Slope Information for Registered Man-made Slopes (2004), 109 p.
Author : Y.F. Yu
Statistical methods and database manipulation have been used to study the correlation between rainfall, landslide frequency and slope information for registered man-made slopes. The approach can be summarised as follows:
- Review rainfall, slope and landslide records.
- Establish databases of rainfall, slope information and landslide records.
- Using a sample of slopes for which comprehensive data are available, establish correlations between rainfall and landslide frequency for different classes of slopes.
- By applying the correlations to the total population of registered slopes in a study area for the period 1996-2000, estimate the number of landslides which would be expected for different slope classes under different rainfall conditions in the area.
- Assess the reliability of the correlations by comparing the number of actual failures with the number of failures estimated from the correlations.
The correlations for soil cut slopes, rock cut slopes, fill slopes and retaining walls appear to follow a semi-log relationship. Slope characteristics such as slope height and angle have a significant effect on the correlations. Based on the correlations, corresponding models have been proposed and used to predict the number of landslides expected for the selected rainstorms in the years 1984 to 2000. The assessment of the expected numbers of landslides shows that the numbers of landslides predicted by the proposed models are reasonably close to the actual numbers of landslides, especially for the soil cut slopes. Uncertainties in the models include errors in the estimation of rainfall intensities by interpolation, and improvement of slopes under the LPM, maintenance and development programmes. It is therefore recommended that the correlations and the landslide prediction models are kept under review.
|1||Contents, Tables, Figures, Appendix (6.49MB)|