246 A Western and R Grayson spatial organisation feature that is not captured by standard geostatistical techniques (variography). We were able to show that indicator geostatistics (indicator variograms) are also unsuitable for characterising connectivity, despite suggestions to the contrary in the literature (Western et al., 1998b). However, connectivity statistics (Allard, 1994, 1993) provide an appropriate statistical tool for characterising spatial connectivity (Western et al., 2000). The use of patterns in model testing is valuable but has some limitations. Many of the simulated patterns were quite similar and it was difficult to assess visually which was the better simulation, especially when patterns for all twelve occasions were considered. There is a need for quantitative pattern comparison techniques that account for a range of different scales including points, hillslopes and catchments. The statistics used for these comparisons need to be chosen carefully so that hydrologically important aspects of the patterns are compared. This might require the comparisons to focus on specific components of the landscape, for example, drainage lines rather than ridge tops, or on correctly simulating pattern features such as connectivity. It is also important that any quantitative approach be able to deal with several different types of data. Here the combined use of soil moisture pattern, saturation deficit time series and runoff time series data was extremely valuable. ACKNOWLEDGEMENTS The Tarrawarra catchment is owned by the Cistercian Monks (Tarrawarra) who have provided free access to their landand willing cooperation throughout the project. Funding for the above work was provided by the Australian Research Council (project A39531077 and S3913076), the Cooperative Research Centre for Catchment Hydrology, the Oesterreichische Nationalbank, Vienna (project 5309), and the Australian Department of Industry, Science and Tourism (International Science and Technology Program).