The importance of management of greenhouse crop transpiration increases, being part of both the water and energy balance. A simulation model for crop transpiration can serve as a soft-sensor in an early warning system for the grower, and is an essential component of an energy model for a greenhouse with a crop. Published data on model validation of crop transpiration under commercial settings are scarce. In an effort to develop a model-based soft-sensor for crop transpiration, continuous and instantaneous rates of crop transpiration were obtained over a large part of 2006 from a tomato grower using a weighing gutter. The wide variation in environmental conditions caused similarly wide variation in crop transpiration rates, both among and within days. This enabled broad model validation. Validation gave over-estimation of crop transpiration, but parameters that relate the stomatal conductance to environmental conditions were successfully calibrated on the basis of total daily transpiration. Although seasonal calibration may in certain cases be sufficiently accurate to enable robust simulation of daily course of crop transpiration, a more robust approach is to calibrate for shorter time periods than an entire season. Robustness is a prerequisite for on-line management of the water and energy balances. Further data analysis can reveal structural patterns in the relations between model parameters and simulated transpiration. Built on this, on-line sensor information on transpiration can be used to continuously optimize the transpiration model, and increase its usefulness in information and early-warning systems.