Modeling plant-soil-water interactions for long periods of time is important for crop management strategies. Because it is difficult to assess soil hydraulic properties, mechanistic models are rarely used for simulating soil water balance. Models less complicated with respect to soil water flow are more often used. Although simple water balance models are easy to use, however, they do not always provide good approximations of plant transpiration and soil-water status. Our objectives were to test the ability of a model, mechanistic with respect to soil-water flow and empirical for soil-plant and plant-atmosphere interactions, to predict soil-water balance components for long periods of time when input parameters are measured or estimated independently. A data set gathered in Nicaragua during several months was used for this purpose. Soil hydraulic properties were measured independently and parameters taken from the literature were used for plant processes modeling. The model predicted reasonably well the soil-water balance for a maize (Zea mays L.)-sorghum [Sorghum bicolor (L.) Merr.] sequence and for a grass sod. Periods of crop water stress were also correctly reproduced. Adjusting some plant parameters slightly increased the goodness of the predictions. Results indicated that plant water uptake was not well predicted when topsoil layers were dry and deep soil layers wet. In addition, the model predicted the upward water flux into the root zone and its important contribution to plant water uptake. During drought periods, the upward water flux reached 2 mm d-1 while the actual evapotranspiration of the crop was between 2 and 4 mm d-1.