Quantitative analysis of multi-temporal SAR datasets requires accurate radiometric calibration. This can usually be achieved by considering a number of multiplicative factors, to be applied to statistically-homogeneous areas on each image to be calibrated. We propose an automated procedure to easily obtain a relative radiometric calibration of a stack of an arbitrary number of ERS SAR images, all coregistered to a unique master. The procedure relies upon the theoretical invariance of both system and scene parameters for targets whose amplitude values are statistically correlated with each other. In practice, with the above-mentioned assumptions, the only varying quantity to take into account is the temporal variation of the calibration factor. We show, based on experiments on two multi-temporal datasets, that the ERS-1 calibration constant exhibits practically no variations except for the expected statistical fluctuations of about /spl plusmn/1 dB. This observation is consistent with the reports on the operations of the ERS-2 satellite, published by ESA. The proposed procedure then consists in applying the full absolute calibration procedure only to the master image, and then relatively correct all other images in the stack by the method described.