Using inverse methods for estimating soil hydraulic properties from field data as an alternative to direct methods

Water and solute transport in the vadose zone greatly depends on me physical and chemical properties of the soil, which generally exhibit high variability. Additionally, the experimental determination of those properties in the field or laboratory is tedious, time-consuming and involves considerable uncertainty for most practical applications. Recently, inverse modeling has been introduced to estimate effective properties in situ by deducing them from, e.g. a measured time series of soil water content. Inverse methods combine forward soil water flow models with appropriate optimization algorithms to find the best parameter set that minimizes an objective function. Global optimization methods are suitable for locating a global optimum for a given set of conditions (number of parameters, boundary conditions, etc.). In this raper we estimate the soil hydraulic properties of a sprinkler fertigated banana plot in the North of Tenerife (Canary Islands) in a direct and inverse way. For the inverse method, use was made of the measured time series of soil water content at three different depths. The forward model is the numerical solution of the Richards equation as implemented in the agro-environmental model WAVE. Two inverse methods are compared: the traditional "trial and error" method and an inverse method using a global search algorithm referred to as the global multilevel coordinate search combined sequentially with the Nelder-Mead simplex algorithm (GMCS-NMS). The global search is shown to be a relatively efficient procedure for estimating the soil hydraulic properties from measured soil water contents inthe banana plantation. However, some ill-posedness problems are identified which should be solved by upgrading the quality of the experimental programs in irrigated banana plantations. (C) 2002 Elsevier Science B.V. All rights reserved.