We examine the evolution of the water production of comet 67P/Churyumov-Gerasimenko during the Rosetta mission (2014 June-2016 May) based on in situ and remote sensing measurements made by Rosetta instruments, Earth-based telescopes and through the development of an empirical coma model. The derivation of the empirical model is described and the model is then applied to detrend spacecraft position effects from the Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA) data. The inter-comparison of the instrument data sets shows a high level of consistency and provides insights into the water and dust production. We examine different phases of the orbit, including the early mission (beyond 3.5 au) where the ROSINA water production does not show the expected increase with decreasing heliocentric distance. A second important phase is the period around the inbound equinox, where the peak water production makes a dramatic transition from northern to southern latitudes. During this transition, the water distribution is complex, but is driven by rotation and active areas in the north and south. Finally, we consider the perihelion period, where there may be evidence of time dependence in the water production rate. The peak water production, as measured by ROSINA, occurs 18-22 d after perihelion at 3.5 +/- 0.5 x 10(28) water molecules s(-1). We show that the water production is highly correlated with ground-based dust measurements, possibly indicating that several dust parameters are constant during the observed period. Using estimates of the dust/gas ratio, we use our measured water production rate to calculate a uniform surface loss of 2-4 m during the current perihelion passage.