DassFlow v1.0: a variational data assimilation software for river flows

Marc Honnorat (1) , Joel Marin (1) , Jerome Monnier (1) , Xijun Lai (2)

1 MOISE - Modelling, Observations, Identification for Environmental Sciences
2 Niglas - Nanjing Institute of Geography and Limnology

Dassflow is a computational software for river hydraulics (floods), especially designed for variational data assimilation. The forward model is based on the bidimensional shallow-water equations, solved by a finite volume method (HLLC approximate Riemann solver). It is written in Fortran 95. The adjoint code is generated by the automatic differentiation tool Tapenade. Thus, Dassflow software includes the forward solver, its adjoint code, the full optimization framework (based on the M1QN3 minimization routine) and benchmarks. The generation of new data assimilation twin experiments is easy. The software is interfaced with few pre and post-processors (mesh generators, GIS tools and visualization tools), which allows to treat real data.

Mots clés: Variational data assimilation; parameter identification; river hydraulics; flood

Domaines: Optimisation et contrôle [math.OC]; Modélisation et simulation; Logiciel mathématique [cs.MS]

Marc Honnorat, Joel Marin, Jerome Monnier, Xijun Lai. DassFlow v1.0: a variational data assimilation software for river flows. [Research Report] RR-6150, INRIA. 2007. ⟨hal-00908169⟩