Low-rank update of preconditioners for the nonlinear Richards equation

Preconditioners for the Conjugate Gradient method are studied to solve the Newton system with symmetric positive definite (SPD) Jacobian. Following the theoretical work in Bergamaschi et al. (2011) [4] we start from a given approximation of the inverse of the initial Jacobian, and we construct a sequence of preconditioners by means of a low rank update, for the linearized systems arising in the Picard–Newton solution of the nonlinear discretized Richards equation. Numerical results onto a very large and realistic test case show that the proposed approach is more efficient, in terms of iteration number and CPU time, as compared to computing the preconditioner of choice at every nonlinear iteration.