Accurately modeling arterial input function (AIF) is important for dynamic contrast enhanced (DCE) MRI. Simulations were performed comparing nine population AIF models to the Parker AIF. Effects of AIF second pass with and without adding noise onto extracted physiological parameters were evaluated with n=1,000 randomly generated physiological parameters (Ktrans and ve) used to calculate contrast agent concentration curves using the Tofts model and Parker AIF. Results demonstrated that the six-parameter linear function plus bi-exponential function AIF model was almost equivalent to Parker AIF. Effects of the second pass were small, unless noise with signal-to-noise ratio was <10 dB.