With the shift of the safety standards for levees from exceedance probability (standard 1996) to flooding probability (new standard), the part of the levee failure process after the initial occurrence of a failure mechanism becomes more important for the calculation of the strength and thus, reliability of the levee cross-section. Interactions between failure mechanisms can take place in this part of the levee failure process. An interaction refers to one failure mechanism influencing the probability of another failure mechanism, where positive interaction is defined as a reduction and a negative interaction as an increase of the failure probability. The main problem is a lack of knowledge about the interactions of failure mechanisms and their influence on the failure probabilities of levees. Several studies (Calle, 2002; ’t Hart et al., 2016; Kok et al., 2017) suggest that there is an interaction between different failure mechanisms. However, only the effects of overtopping on slope instability (de Visser et al., 2018) and a reduction of the shear strength in slope instability due to uplift (Kanning and van der Krogt, 2016) is currently included in the statutory assessments of levees (Rijkswaterstaat WVL, 2017a,d). This research aims to quantify the influence of the interaction between the failure mechanisms backward erosion piping (BEP) and slope instability (SI) on the safety of a levee. The parametric study of this research assesses whether the occurrence of a failure mechanism (BEP or SI) affects the parameters' present states in the limit state functions of the other failure mechanism. A model was created to calculate this interaction between SI with BEP. After a slope instability, the remaining profile is assumed to contain a berm. This change in levee profile, does not result in the immediate breaching of the levee, it can however influence the probability of BEP. The model sequentially executes a stability analysis, a displacement analysis of the sliding plane and a BEP analysis. The degree of interaction is defined as the difference between the safety factor of BEP of the original profile (prior to sliding) and the safety factor of BEP of the remaining profile (after sliding). The modelling study shows that the degree of interaction from SI to BEP is significantly affected by the presence of cracks in the berm. A consequence of these cracks is that the exit point moves and is now located in the berm. The degree of interaction also depends on the shape and size of the sliding plane, because it determines the size of the berm of the remaining profile. A vast majority (50\% to over 75\%) of the considered scenarios of the seven considered levee profile cases result in a positive interaction. However, in all the considered cases, negative interactions were found that give a reduction of 50\% to 2\% on the original safety factor, which could lead to an overestimation of the safety in the assessment of the levee. All of these negative interactions result from the assumption that large cracks can form in the soil after sliding.

The Addicks and Barker Reservoirs, built in the forties, are located in Houston and collect precipitation and run-off from upstream areas to reduce flood risks along Buffalo Bayou to protect downtown Houston. During Hurricane Harvey (August 25 - August 30, 2017), the precipitation reached a new record of 910 mm [36.2 inches] in a 4 day period in Houston. The gates of Addicks and Barker Reservoirs were opened during the night of 27-28 August which led to major damages due to downstream flooding. Besides, non-government owned land upstream was flooded due to high water levels in the reservoirs.
In this report, new design water levels for Addicks and Barker Reservoir are calculated based on inflowing discharge into the reservoirs and precipitation directly onto the reservoirs, including data of Hurricane Harvey. These calculated design water levels are compared with the critical water levels calculated based on the failure mechanisms of the dams. This study shows that the original design water level of the dams, based on the Probable Maximum Flood, are 2.83 m and 1.01 m higher than the critical water level for which failure of the dams can occur due to piping for Addicks and Barker Reservoir. However, the maximum allowed water level which is currently maintained by the United State Army Corps of Engineers, is 2.19 m and 2.46 m below the calculated critical water level. During Hurricane Harvey, these maximum allowed water levels were exceeded with 3.46 m and 1.93 m.
The damage of residential properties upstream and downstream of the reservoirs are minimized based on the distribution of excess volume from the inflow of creeks and precipitation onto the reservoirs. The ratio of the amount of volume which should remain upstream of the dams and the volume discharged into the Buffalo Bayou is calculated for every considered event with its duration and return period. The ratio of Addicks Reservoir is the dominant ratio, which should be used for both reservoirs. Run-off alone already produces damage, especially for the 12h and 24h precipitation, so the Addicks and Barker Reservoirs should not release discharge into the Buffalo Bayou for small durations. For events with a longer duration, it would cause less damage to open the outlets of the reservoirs than to keep them closed. However, if the water level in the reservoir exceeds the critical water level for piping, it is advised to discharge more to the downstream area to prevent breaching of the dams. Since the critical water level is reached for approximately 25% of the events at Addicks Reservoir, mitigations against piping should be taken to improve the minimization of damage. For Barker Reservoir, the critical water level is not reached in the optimization. During big events, people living upstream will be more affected by the flooding than people living downstream since this optimization is based on the damage minimization of residential properties.