Estimation of levee breach discharge hydrographs: comparison of inverse approaches

In this paper, we present a methodology to estimate the flow hydrograph through a levee breach in an inverse framework. The procedure requires a forward model of the river reach, water levels recorded in a gauging station downstream or upstream of the failure site and an optimization method; there is no need to model either the flood-prone area outside the river or the breach evolution. In particular, we compare the results of a novel optimization method, the ensemble smoother with multiple data assimilation (ES-MDA), with those of the Bayesian geostatistical approach (BGA). The methods are applied to the real case of the breach that occurred on the Secchia River in Northern Italy in January 2014. The two optimization methods lead to similar results, but ES-MDA requires a number of runs of the forward model about 10 times lower than that needed by BGA, saving significant computational time.