Erodible river banks need to be safe against the possible scouring during flood events in mountain rivers. The consequences of the bank failure are probably lateral uncontrolled erosion and flooding with disastrous losses in residential areas or damage of infrastructures. Among all flood protection measures that keep the riverbank safe against erosion and damage, riprap revetment is one of the most commonly used structures. In order to ensure the safest design, determining the required riprap size is one of the most important factors. Several methods of riprap sizing exist which are mostly evolved for dumped median size blocks. However, in mountain rivers and steep channels, the extra stability has to be ensured by using the large, heavy blocks as bank riprap protection, which have to be individually placed because of their weight. Such arrangement generates additional resistance against flow erosion since the space between the blocks is minimized, and the interlocking forces among the blocks are increased. The behavior of the latter protection was rarely studied for alpine river conditions, and no adapted design criteria exist. Therefore, an experimental study was carried out. This research investigates the effect of packed placement of riprap on sizing, the resistance to failure and the time to failure of riprap. Comparison with the existing design methods is also performed considering the effect of riprap thickness and bank slope. This is studied by means of 123 series of systematic tests of compressed riprap and 34 tests of dumped ones. The experiments were carried out using a 10 m long and 1.5 m wide flume with a rough fixed bed at the Laboratory of Hydraulic Constructions (LCH) at École Polytechnique Fédérale de Lausanne (EPFL). Riprap was reproduced with uniform crushed stones with three block sizes namely D50= 0.037, 0.042 and 0.047 m. Tests were conducted on streamwise bed slopes of S= 0.015, 0.03 and 0.055, and riprap bank slopes of 35°, 31° and 27° under supercritical flow conditions and for a maximum of three hours of test duration. The porosity which is an effective factor on the stability is assessed and the results show a reduction of 2% for smallest size to 10% for the largest size of packed blocks. An empirical relationship between relative roughness and modified Froude number is discussed. Then a sizing riprap empirical formula for large blocks individually placed on supercritical flow is herein developed; considering the riprap thickness and bank slope. This empirical relationship is compared with existing formulae. As a further step based on a time-based analysis of the failure process, a relationship among time to failure, shear velocity, and dimensionless bed shear stress is established. An empirical relationship was established which allows to estimate the time to failure of the riprap bank protections. The influence of a second layer on the time to the failure and on the bank stability is also analyzed. For the same longitudinal channel slope and bank slope, the second layer significantly stabilizes the bank protection and postpones its failure time. Nevertheless, during the test, the block erosion rate is increased significantly (almost twice) for two layers comparing to one layer riprap. Finally, the potential failure probabilities of riprap are evaluated by using Monte Carlo Simulation and Moment Analysis Methods as well as Rosenblueth Point Estimation Method. The advantages of this probabilistic model are that it can cover different failure mechanisms and make use of the multivariate probabilistic method. The probability of failure in various modes, namely direct block erosion, toe scouring and overtopping, has been defined. The method was applied to two rivers in Switzerland; namely Kleine Emme and Brenno. The probability of failure for different mechanisms based on the expected sediment transport regime under climate change is defined for these two rivers as a case study.