Prediction of the influence of waste application on agricultural land and on the dynamics of infiltration is crucial for the optimum management of soil water as well as contaminants from runoffs. Three models (Philip’s, Kostiakov’s, and Horton’s) were investigated for their capability to describe water infiltration into a Typic Haplustult amended with different rates 10, 12.5, 25.0, 37.5 and 50.0 Mg ha-1of fresh (FW) and burnt (BW) rice-mill waste. Data were collected for two seasons between 1991 and 1992. Based on the values of the coefficient of correlation (R), the Kostiakov’s model provided the best fit with experimental data for both fresh (FW) and burnt (BW) rice-mill waste for the two seasons. It was followed by the Philip’s and then Horton’s models. However, transmissivity coefficients (A) of the Philip’s model were negative while Kostiakov’s coefficients were very insensitive to variations in application rates (q) of waste. Since the Horton’s coefficients indicated the highest sensitivity to q, these coefficients were expressed in terms of q and then used for the prediction of cumulative infiltration. Variation in these coefficients with q were exponential and parabolic with R2 ranging from 0.867 to 0.891 and 0.623 to 0.783 for the FW and BW amendments, respectively. Incorporation of q increased R2 from the poor negative average value of-0.382 to 0.748, thereby providing tools for advance prediction and analysis without actual waste application.