Application of finite-difference methods to parameter identification of agroecological models

Abstract

A typical parameter identification problem arisen in agroecological modeling is described and a finite difference method to solve an appropriate inverse problem is proposed. The method sup­poses the direct solution of an initial-boundary problem by finite difference scheme and then uses a non-linear optimization routine for parameter estimation. The derived recurrent formulas allow the accumulation of values of gradient and Hessian matrices while solving the direct problem. The software algorithm is described and some numerical results are given for the particular soil heat transfer parameter estimation exercise using the field treatment data collected at the Lithuanian Institute of Agriculture.