HYDROLOGICAL PROCESSES, VOL. zyxwvu 9, 445-467 (1995) PREDICTING CATCHMENT-SCALE SOIL MOISTURE STATUS zy WITH LIMITED FIELD MEASUREMENTS J. D. KALMA zyxwv CSIRO, Division zyxwvutsr of Water Resources, Canberra Laboratory, GPO Box 1666, Canberra, ACT 2601, Australia B. C. BATES CSIRO, Division of Water Resources, Perth Laboratory, Private Bag, PO Wembley, WA 6014, Australia AND R. A. WOODS Centre .for Water Research, The University of Western Australia, Nedlands, W A 6009, Australia ABSTRACT A catchment-wide soil moisture index based on spatially distributed point measurements of soil moisture is used to describe the temporal trend in regional soil moisture status in a 26 km2 catchment in south-eastern Australia. The tem- poral variation in runoff, evaporation and soil moisture storage is simulated with a modification of the lumped SFB water balance model of Boughton (1984), which assumes a fixed bucket size, and with the variable infiltration capacity (VIC) model of Wood et al. (1992), which assumes a variable bucket representation. Comparison of simulated catch- ment soil moisture storage and the soil moisture index based on measurements indicates that both models can make useful predictions of soil moisture status at the catchment scale, with the VIC model performing slightly better than the SFB model. It is also shown that the quasi-distributed VIC model can predict the relative wetness at individual locations, given their relative frequency of occurrence, thus allowing the disaggregation of catchment-scale storage values to point-scale soil moisture values. KEY WORDS Lumped water balance models Soil moisture measurements Catchment hydrology Scale INTRODUCTION Lumped water balance models use simple one-dimensional representations of hydrological processes to provide catchment-wide estimates of the water balance components. Such models use spatial averages of precipitation, soils, geology or topography. The spatial averaging implies that the whole catchment can be represented mathematically using only the dimensions of time and depth (Eeles et al., 1990). Many regional and macro-scale studies of the hydrosphere still rely on simple ‘bucket’ models, although there is now increasing effort towards accounting for spatial variability in land surface parameters. This may be achieved by assuming statistical distributions for key (sub)surface parameters (e.g. Entekhabi and Eagleson, 1989; Wood et zyxwvut al., 1992) or by disaggregating regions, basins and global circulation model (GCM) grid squares into more homogeneous subregions to reduce spatial variability in land surface parameters. For example, Jolley and Wheater (1993) used the MORECS system (Thompson et al., 1981) to make water balance calculations for each major land-use class present in 40 zyxwvutsrq x 40 km grid squares covering the 10 000 km2 Severn River catchment. Fluxes and soil moisture deficit values are averaged according to land-use fractional areas, similar to the method proposed by Avissar and Pielke (1989). Lumped rainfall-runoff models are usually highly parameterized. They require calibration using CCC OS85-6087/95/030445-23 zyxwvutsrq 0 1995 by zyxwvutsrq John Wiley & Sons, Ltd. Received 15 April 1994 Accepted 14 December 1994