The frequent occurrences of urban floods around the globe have been mainly attributed to the improperlydesigned parameters of urban stormwater drainage sewers and the unexpected occurrence of extreme rainfall.It is expected to increase the number and frequency of future extreme rainfall eventsunder a warming climate, which increases the pressure on the carrying capacity of the urban stormwater sewers leading to frequent failures. Conventionally, the design storm approach used for sizing stormwater sewers is deterministic and relies on two major assumptions: (i) average recurrence interval neutrality and (ii) adequacy ofcritical rainfall duration (i.e., time of concentration) fordesign flood estimation.However, both assumptions are questionable as thereturn periods of rainfall and runoff may not be equal, and a single critical duration may neglect the complete time profile of the rainfall within an extreme event. The designs are usually carried out for a specific rainfall return period (e.g., 2-years) from the historical observations, and considering stationarity in climate, the structures are assumed not to fail more often than the design return period in the future.Further, the current design procedures do not address the consequences(flood risk and vulnerability) of the failure of a structure when it does fail.
In this study, we propose a safe-fail approach to design urban stormwater drainage infrastructures and demonstrate its effectiveness on a small urban catchment of Chennai city, India. The approach is developed to meet the failures withminimal consequences of the flood during a failure. The procedure involves two steps:(i) a risk (probability of failure) based initial design,and (ii) a flood consequences based updation of designs. In the first part,all the storm sewers were designed to carry a 2-year return level discharge. The estimation of such return levels involved development of a SCS based hydrologic model for catchment outlet discharge estimation, a new threshold selection procedure to identify extreme rainfall events, and a new desegregation technique to obtain sub-hourly (15-min) rainfall pulses from hourly magnitudes.The return periods of each link were updatedto minimize the consequences of flood in the later step through an ensemble simulation of flood depthsand an iterative procedure.Thesteps involved were,ensemble (1000-years) generation of extreme rainfall events using a coupla based bivariate frequency analysis, development of a Markov chain based stochastic model to characterize the time profile of extreme rainfall events,and simulation of flooding consequences using a combination of 1D-2D HEC-RAS flood inundation model.Further, extreme rainfall projections from 14 GCMs of CMIP6 project were bias-corrected using a novel L-moments scaling method to account for the possible future climate change scenarios in the current design.The proposed safe-fail approach facilitates the development of a relationship between flooding consequences (flood depth) and design return period of storm sewers, which assists in the selection of appropriate return period for each storm sewer such that an inevitable failure of sewers is safe, as demonstrated in this work.