In greenhouses, reducing water consumption by increasing water efficiency is of high interest. To reach this goal, predictive models of soil-plant-atmosphere water transfer could be helpful. However, such models have been mainly developed for open field conditions, and very few models exist for greenhouse plants grown in pots. Moreover, most of these models were implemented under well-watered conditions, but very few are available under water restriction. The aim of this study is to develop an integrated soil-plant-atmosphere water balance model applied to potted plants grown in greenhouses, to predict plant transpiration under different restrictive irrigation regimes. Implementing such a model requires an accurate estimation of stomatal resistance Rs under water restriction conditions. Rs is then used in the Penman-Monteith model to evaluate transpiration. To establish the model parameters, an experiment was conducted for sixteen weeks inside a greenhouse with ornamental plants (New Guinea Impatiens) grown in containers on shelves. Well-watered and water restriction conditions were applied. The peat matric potential, radiation, temperature and humidity were continuously recorded, while Rs was measured and transpiration was assessed every half-hour from 8:00 am to 8:00 pm 10, 11, 12, 14, and 16 weeks after planting. The resulting model was first validated against experimental measurements during the twelfth week of the experiment. It displayed good correlations for both the instantaneous data and integrated total transpiration. Different scenarios of irrigation reduction (frequency x volume) were tested and the results indicated real potential for water use reduction. Indeed, by reducing water application by 50% and by applying one irrigation per day, transpiration was not affected