Four distinct approaches, that vary markedly in the spatial and temporal resolution of their measurement and process-level outputs, are used to investigate the daily and seasonal water vapour exchange in a 70-year-old Belgian Scots pine forest. Transpiration, canopy interception, soil evaporation and evapotranspiration are simulated, using a stand-level process model (SECRETS) and a soil water balance model (WAVE). Simulated transpiration was compared with up-scaled sap flow measurements and simulated evapotranspiration to eddy covariance measurements.Reasonable agreement in the temporal trends and in the annual water balance between the two models was observed, however daily and weekly predictions often diverged. Most notably, WAVE estimated very low, to no transpiration during late autumn, winter and early spring when incident radiation fell below ,50 W m 22 while SECRETS simulated low (0.1-0.4 mm day 21 ) fluxes during the same period. Both models exhibited similar daily trends in simulated transpiration when compared with sap flow estimates, although simulations from SECRETS were more closely aligned. In contrast, WAVE over-estimated transpiration during periods of no rainfall and under-estimated transpiration during rainfall. Yearly, total evapotranspiration simulated by the models were similar, i.e. 658 mm (1997) and 632 mm (1998) for WAVE and 567 mm (1997) and 619 mm (1998) for SECRETS.Maximum weekly-average evapotranspiration for WAVE exceeded 5 mm day 21 , while SECRETS never exceeded 4 mm day 21 . Both models, in general, simulated higher evapotranspiration than that measured with the eddy covariance technique. An impact of the soil water content in the direct relationship between the models and the eddy covariance measurements was found.The results suggest that: (1) differentmodel formulations can reproduce similar results dependingon the scale at which outputs are resolved, (2) SECRETS estimates of transpiration were well correlated with the empirical measurements, and (3) neither model fitted favourably to the eddy covariance technique. q