PAREMERA, implemented in Fortran90, uses the generalized Gauss–Newton method to compute the solution of the parameter estimation problem. Thus, it has the advantage to not converge towards to instable local stationary points and is stable against small perturbations in measurements. Furthermore, PAREMERA decomposes the time interval using multiple shooting and thus reduces the nonlinearities of stiff systems. It further implements the so-called reduced approach that allows solving parameter estimation problems with PDE or high-dimensional ODE constraints in acceptable time. PAREMERA is particularly outlined to be used in combination with a software tool for algorithmic differentiation, which provides exact derivatives (up to machine precision) of the model functions. Bindings to the open source FEM deal.ii, which is widely used in the community of scientific computing, are available.

[1] Robert Kircheis: Structure Exploiting Parameter Estimation and Optimum Experimental Design Methods and Applications in Microbial Enhanced Oil Recovery, PhD Thesis, Heidelberg University, 2016. Online. http://archiv.ub.uni-heidelberg.de/volltextserver/22098/