We describe the calibration, measurements and data reduction, of the dark c
urrent of the ISOCAM/LW detector. We point-out the existence of two signifi
cant drifts of the LW dark-current, one throughout the ISO mission, on a ti
mescale of days, another within each single revolution, on a timescale of h
ours. We also show the existence of a dependence of the dark current on the
temperature of the ISOCAM detector. By characterizing all these effects th
rough polynomial fittings, we build a model for the LW calibration dark, th
at depends on the epoch of observation (parametrized with the revolution nu
mber and the time elapsed in that given revolution since the activation) an
d on the temperature of the ISOCAM detector. The model parameters are tuned
for each of ISOCAM/LW pixel. We show that the modelling is very effective
in taking into account the dark-current variations and allows a much cleane
r dark subtraction than using a brute average of several calibration dark i
mages. The residuals of the LW model-dark subtraction are, on average, simi
lar to the pre-launch expectation.