The sensitivity of a catchment model to soil hydraulic properties obtained by using dierent measurement techniques Sharon H. Davis,* Robert A. Vertessy and Richard P. Silberstein CRC for Catchment Hydrology, CSIRO Land and Water, PO Box 1666, Canberra, A.C.T. 2601, Australia Abstract: Most studies on the use of physically based hydrological models have identi®ed saturated hydraulic conductivity (K sat ) as one of the most sensitive input parameters. However, K sat is also one of the most dicult landscape properties to measure accurately, casting doubt on the ability of modellers to estimate this parameter a priori for catchment simulations. Several studies have shown that K sat estimates are greatly in¯uenced by the measurement method used, primarily because of scale eects. In this paper, we evaluate the eect of K sat measurement method on catchment simulations aimed at predicting water yield from forested catchments. A series of simulations are conducted using the Topog Dynamic catchment model, with K sat estimated by means of the constant head well permeameter, small core 63 cm  73 cm and large core 223 cm  30 cm methods. These were applied in a deep, permeable forest soil in which macropore ¯ow has been noted to occur. The three measurement methods yielded very dierent K sat estimates and these had a large eect on model results. The model predictions based on small core and well permeameter measurements were extremely poor, as these methods did not adequately account for preferential ¯ow through the soil. The large core estimates of K sat , which were one to three orders of magnitude higher than the values obtained by the other two techniques, produced good predictions of catchment discharge and known spatial patterns of water table depth. Our results highlight the need for caution when applying soil hydraulic measurements to catchment-scale models. Copyright # 1999 John Wiley & Sons, Ltd. KEY WORDS saturated hydraulic conductivity; catchment models; Topog Dynamic INTRODUCTION It is well recognized that in order to test physically based catchment models accurate input data are required. To date, our capacity to develop complex models has progressed more rapidly than our ability to measure soil hydraulic properties accurately in the ®eld (Clothier and Smettem, 1990; Hillel, 1991; Vertessy et al., 1993). These problems are most pronounced in the measurement of the saturated hydraulic conductivity (K sat ) of a ®eld soil, owing to the heterogeneous and anisotropic nature of many soils and the contribution of soil structural features to water movement under saturated conditions. In addition, the spatial distribution of soil hydraulic properties makes catchment-wide parameterization dicult. It has been argued that diculties in accounting for spatial variation limit the utility of physically based catchment models (Loague and Freeze, 1985). There is also debate in the literature regarding the validity of extrapolating point measurements to the element scale used in physically based models (Beven, 1989; Grayson et al., 1992). However, this CCC 0085±6087/99/050677±12$1750 Received 21 January 1998 Copyright # 1999 John Wiley & Sons, Ltd. Revised 5 June 1998 Accepted 11 August 1998 HYDROLOGICAL PROCESSES Hydrol. Process. 13, 677±688 (1999) *Correspondence to: S. H. Davis, CRC for Catchment Hydrology, CSIRO Land and Water, PO Box 1666, Canberra, A.C.T. 2601, Australia.