STABILITY OF RESERVOIR RIM SLOPES SUBJECTED TO FLUCTUATING WATER LEVEL

Abstract

Existence of the reservoir-induced landslides has been acknowledged worldwide as a major barrier in the effective and successful functioning of hydro-power projects. Events of reservoir induced landslides have simultaneously risen in the past few decades along with the increase in number of hydro-power projects. Reservoir-induced landslides are more complex in nature than the landslides observed in mountainous regions as there can be several causes of failure. The triggering factors for the reservoir-induced landslides are mainly seepage, rainfall, geometric instabilities, and reservoir-water level fluctuations. Reservoir-induced landslides are often linked to fluctuations in reservoir water levels. For instance, more than 5000 landslides have been witnessed in Three Gorges Reservoir (TGR) since the commissioning of the TGR Dam, majority of which have been attributed to water level fluctuations. There exist various theories that explain the failure mechanisms of these landslides due to such fluctuations. Still, the failure mechanism of reservoir-induced landslides due to water level fluctuations is not clearly understood. Besides, limited studies are available on the displacement prediction models for rim slopes experiencing movements implicitly due to reservoir level fluctuations. The available prediction models are based on long-term in-situ field displacement monitoring. A detailed investigation is essential to understand the various aspects of the failure mechanism of these fluctuations-induced landslides. The contribution of reservoir water level fluctuations as an isolated triggering factor needs to be quantified. With the increasing number of hydro-power projects in India as well as throughout the globe, a simple quick assessment tool for predicting rim slope displacements based on easily available field slope data is the need of the hour. The present study aims at exploring the failure mechanism of rim slopes specifically due to reservoir water level fluctuations, and to come up with a simple generalized empirical solution to predict the displacements along the field rim slopes based on available field data. To achieve the objectives of the present study, an extensive experimental program was planned followed by seepage and stability analyses of the test results. A small-scale physical slope model test (PSMT) apparatus was developed indigenously to study the rim slopes subjected to fluctuating water levels. Eight PSMTs were performed in total. Two PSMTs were performed in steady-state condition to study the initiation and propagation of rim slope failures in steady-state seepage i condition with varying soil type (c-ϕ soil from Koteshwar, and sandy soil from Solani river) and slope inclination (30° and 45°). Six PSMTs were performed on a modelled soil (i.e. scale-reduced soil derived from a rim slope site in Koteshwar reservoir using parallel gradation technique) subjected to reservoir water level fluctuations. The six PSMTs were performed by varying cyclic fluctuation rate (i.e. 16 cm per 6 hours – slow, 16 cm per 1 hour – moderate, and 16 cm per 10 minutes - rapid), slope inclination (30° and 45°), and number of cycles (i.e. up to 30 cycles). During the experimentation, total head values with the help of piezometers installed on the setup, degradation of slope face, and deformed slope profiles with the help of deformation gauge, were observed at different test stages. The observations were made until the model slope failure occurred in steady-state condition, and up to the completion of 30 cycles when subjected to reservoir fluctuations. Lastly, seepage and stability analyses of the PSMTs results were performed to gain deeper understanding about the failure mechanism. The results and analyses of PSMTs in steady-state condition revealed that the cause for initial failures in the case of slopes that are reasonably stable under steady-state seepage condition is highly likely due to high exit hydraulic gradient. The initial slope failure will be induced within the region of high hydraulic gradients. The depth of failure surface will mostly be governed by the soil type. As the soil changes from cohesionless soil to a cohesive soil, the depth of failure slip surface tends to be increasing with cohesion being mobilized. There is a high probability of retrogressive failure phenomenon in the case of cohesionless granular soils. Using the experimental observations of PSMTs subjected to cyclic fluctuations, the analyses of PSMT observations were carried out in various ways. Firstly, spread (displacement at toe) was assessed. It was found that the amount of spread is high for slopes subjected to slow cyclic fluctuations in comparison to slopes subjected to rapid cyclic fluctuations. Change in pore water pressure just after the drawdown with successive cycles were studied for the performed PSMTs. The results suggest that sharp accumulation of pore pressure occurs during the initial cycles of water level fluctuation. The rate of increase in pore pressure decreases afterwards, and stabilizes for the remaining cycles of water level fluctuations. Seepage analysis was carried out to validate the trend of total head values observed during experimentation for reservoir water level fluctuations. The pore-pressure derived from the seepage analyses were then used for stability analysis. It was found that the stability of slopes subjected to cyclic fluctuation strongly follows the water level fluctuation scheme. Lastly, volumetric evolution, and static liquefaction ii were evaluated. Static liquefaction assessment concluded that the rim slopes are susceptible to static liquefaction under the influence of cyclic fluctuations. Based on the test observations, deformational behavior of rim slopes subjected to cyclic fluctuations was suggested. It was elicited that deformations and pore-pressure are inter-related to each other. Deformations occur mostly due to the high excess pore pressure generated after drawdown of each cycle. The rate of deformation reduces during the impounding process. Permanent deformations get accumulated with every passing cycle. Initial cycles of water level fluctuations cause large deformations at a high rate with the possibility of shallow toe failure during the initial cycle drawdowns. The rate of deformation reduces over passage of successive cycles. However, repeated cyclic fluctuations induce weakening effect in the slope mass (strain softening phenomenon). Spread observed from the PSMTs was then used to propose a novel approach to predict displacements along rim slopes due to reservoir water level fluctuations. A dimensionless term called Spread Index (Silab) was defined. The index was defined as the ratio of Spread (horizontal displacement at toe) and Hydro-Fluctuation Belt (HFB) of the cyclic fluctuation in the reservoir water level. Spread index was obtained for all the tests subjected to 30 cycles. Spread index data were used to perform multiple linear regression analysis. The spread index (Silab) was considered to be dependent on slope inclination (θsm), cyclic fluctuation rate (CFR) and number of cycles (Nc). The obtained regression equation was then statistically examined for goodness of fit, multicollinearity among the parameters, hypothesis testing by ANOVA. A correction factor Ci, field was defined to account for the influence of change in the ratio of hydro-fluctuation belt and slope height. Finally, an empirical solution to compute the spread in the field was suggested. The proposed empirical solution was validated by assessing the displacement for three landslide sites. The predicted displacements were compared with the already observed displacements in the field. An error of approximately 10% was observed. Lastly, the proposed empirical solution was used to predict displacements along a rim slope situated in Koteshwar reservoir, India.