Modelling transpiration of greenhouse gerbera (Gerbera jamesonii H. Bolus) grown in substrate with saline water in a Mediterranean climate

•Greenhouse soilless gerbera was grown in autumn and spring with fresh or moderately saline water.•The Penman–Monteith model and two regression equations, which responded to vapour pressure deficit and/or radiation, were used to predict the hourly transpiration rate.•Water salinity affected crop growth and production only in spring, but it influenced neither stomatal resistance nor model calibration.•The models explained 80–96% of the variability of measured transpiration. Gerbera plants were grown in semi-closed rockwool culture under greenhouse conditions in different seasons in a Mediterranean climate. The plants were irrigated using either fresh (FW; 1.0molm−3 NaCl) or moderately saline (SW; 9.0molm−3 NaCl) water. In autumn, NaCl concentration did not influence significantly plant growth, flower production and transpiration (E), which instead were reduced in spring in the plants irrigated with SW. In both seasons, water salinity did not affect leaf stomatal resistance (rl), which was determined by the inversion of the Penman–Monteith (PM) equation or measured with a diffusion porometer. The PM formula and two regression equations were calibrated and validated for estimating the hourly rate of daytime transpiration (Ed); a regression model was also fit to nocturnal transpiration (En). Regression models predicted Ed as a function of vapour pressure deficit (VPD) and/or the radiation intercepted by the canopy. Leaf area index (LAI), which is required by all the equations, was modelled as function of crop thermal time (i.e. growing degree days). The PM model predicted Ed using a constant value of rl. Model calibration and validation were performed using independent datasets. The irrigation with FW or SW did not require a different calibration of transpiration models. Both PM formula and regression equations provided accurate estimates of Ed; fitted equations explained between 80% and 96% of the variance in measured Ed. A linear regression of En against (LAI·VPD) accounted for 92% of measured En.