Context : Comet 67P/Churyumov-Gerasimenko was selected in 2003 as the new target of the Rosetta mission .
It has since been the subject of a detailed campaign of observations to characterise its nucleus and activity .

Aims : Here we present previously unpublished data taken around the start of activity of the comet in 2007/8 , before its last perihelion passage .
We constrain the time of the start of activity , and combine this with other data taken throughout the comet ’ s orbit to make predictions for its likely behaviour during 2014/5 while Rosetta is operating .

Methods : A considerable difficulty in observing 67P during the past years has been its position against crowded fields towards the Galactic centre for much of the time .
The 2007/8 data presented here were particularly difficult , and the comet will once again be badly placed for Earth-based observations in 2014/5 .
We make use of the difference image analysis ( DIA ) technique , which is commonly used in variable star and exoplanet research , to remove background sources and extract images of the comet .
In addition , we reprocess a large quantity of archival images of 67P covering its full orbit , to produce a heliocentric lightcurve .
By using consistent reduction , measurement and calibration techniques we generate a remarkably clean lightcurve , which can be used to measure a brightness–distance relationship and to predict the future brightness of the comet .

Results : We determine that the comet was active around November 2007 , at a pre-perihelion distance from the Sun of 4.3 AU .
The comet will reach this distance , and probably become active again , in March 2014 .
We find that the dust brightness can be well described by Af \rho \propto r ^ { -3.2 } pre-perihelion and \propto r ^ { -3.4 } post-perihelion , and that the comet has a higher dust-to-gas ratio than average , with log ( Af \rho / Q ( H _ { 2 } O ) ) = -24.94 \pm 0.22 cm s molecule ^ { -1 } at r < 2 AU .
A model fit to the photometric data suggests that only a small fraction ( 1.4 % ) of the surface is active .

Conclusions :