Round-off error occurs when numerical calculations are approximated due to the finite precision of computer arithmetic, leading to small discrepancies between the calculated and true values. This error is particularly significant in iterative processes and can accumulate, potentially causing substantial deviations from expected results in scientific computations.