Simulation of the water balance of a small 0.05-ha field of dryland rice (grown on non-submerged soil) requires knowledge of the soil's hydraulic conductivity functions. We compared three methods to generate such functions: direct measurement in the laboratory on samples from soil horizons occurring within the field (Method 1); use of average hydraulic conductivity functions measured in a 50-ha dryland rice region surrounding the field (Method 2); and deriving the hydraulic conductivity function through model calibration (Method 3).Differences in hydraulic conductivity functions obtained by the three Methods were large. The impact of these differences on simulation results using a soil-water balance model was evaluated. Simulated and observed water contents in 0 to 0.4-m depth were compared for two representative 0.05-ha fields within the 50-ha dryland rice area during three monsoon seasons and one dry (non-monsoon) season. Good agreement between observed and simulated soil water contents was obtained with input derived from either of the three Methods. Best results were obtained with Method 1, but differences in model performance were small because simulation results were relatively insensitive to hydraulic conductivity input data. Implications for sampling strategies are discussed.