Spatial and temporal variation in two rainfall simulators: implications for spatially explicit rainfall simulation experiments

Rainfall simulators are widely used yet there is little evidence in the literature to show that their spatial and temporal variability has been adequately taken into account. For experiments that are concerned only with some aggregate or mean effect of simulated rain then such variations may be unimportant. However, where rainfall simulation is being used to study (and perhaps model) small-scale processes that are themselves spatially variable (such as rill initiation) then knowledge of the simulator's inherent variability is vital. A first aim of this paper is therefore to examine this variability, and to appraise methodologies by which it may be quantified. A second aim is to evaluate the implications for spatially explicit rainfall simulation experiments. Two simulators were used, a portable drip-screen simulator and a laboratory-based full-cone nozzle simulator. Neither produced a spatially uniform distribution of rainfall depth: both produced distributional patterns that were fairly consistent despite varying intensities and run times. Small-scale, apparently random variations were superimposed on these more deterministic patterns. However, despite this marked spatial variability, calculation of uniformity coefficients (1−SD/mean) resulted in high values. Thus it appears that the uniformity coefficient gives little real indication of the spatial uniformity of simulated rainfall, despite its established usage in the literature. Additionally, spatial distributions of raindrop size –and hence kinetic energy –were calculated for the full-cone nozzle simulator. These show that zones of high rainfall amount do not necessarily relate to zones of high energy reaching the surface. The presence of such variability raises a number of issues for spatially explicit rainfall simulation experiments. While there has been little work on the spatial variability of natural rainfall at field scale and smaller, it appears that the spatial heterogeneity of simulated rainfall depths observed in this study does not differ greatly from that of natural rain. But since a major attraction of rainfall simulation experiments is additional control over rainfall's many variables, the spatial non-uniformity of depth observed in this study is unwelcome. The existence of an apparently deterministic component to this non-uniformity nonetheless suggests that it can, at least in principle, be corrected by calibration. Less easily handled is the discrepancy between spatial distributions of rainfall depth and energy, since this will certainly affect rainfall simulation experiments that are, for example, concerned with erosion processes due to raindrop impact. Copyright © 2000 John Wiley & Sons, Ltd.