The variability of options, points of view and requirements attached to the choice of one of the numerous available tools for crop management, leads to difficulties in technical exchanges among model users. Difficulties arise from the choice of the output and from the different representations of mechanisms adopted by the different users. The approach currently being explored by the CIRAD Cotton Program is to develop new interfaces consistent with user representation as output of the complex crop model COTONS. To illustrate this approach, an example is presented showing soil representation involved in the water balance. Tensiometers were chosen to represent the grower's point of view to schedule crop water requirements. Modeller's point of view is a two compartment representation as adopted in many models compared to a 2D soil representation as in COTONS. By definition, an interface does not change the way phenomena are simulated, but how the output is presented. In that sense, no modification was done to the water balance processes such as evapotranspiration, root growth rate, runoff and percolation that are all simulated by COTONS. The task was then limited to calculate the fraction of available soil water content (FTSW) as proposed by Sinclair and Ludlow (1986) in accordance to the cell and the layer soil representations and to propose an empirical method to link the modelling output to the observed soil water depletion. To deal with this problem, it was proposed to consider a minimum weight of root per cell as limit of the effective rooting depth (ERD). While FTSW is widely used in modelling to quantify soil and plant water status (Sadras and Milroy, 1996), the output water stress index proposed by COTONS (WSTRS) depends on mean cell water potential during the day, the average daily air temperature and the net radiation. As a result of the work being done, comparisons revealed little difference between FTSW calculated in accordance to the 2 compartments and soil