The unrepeatable river: Exploring chaotic variability in laboratory channels

Experimental work relies on the assumption that, under controlled conditions, repeat trials will yield consistent outcomes. This assumption is crucial in geomorphology, as both a theoretical and applied science, yet it is rarely examined directly. We investigated the repeatability of channel forms by conducting 14 trials using a gravel-bed stream table. Analyses of morphology and hydraulics using simple statistics, dimensional reduction techniques and hierarchical clustering were used to quantify similarities and differences between outcomes. Average channel form was consistent across all trials, but practically, irreducible variations in initial conditions led to divergent outcomes through chaotic dynamics. Notably, two distinct channel outcomes were observed, resembling attractor states, relating to the presence or absence of an avulsion. The results indicate that even under well-controlled conditions, repeatability is not guaranteed in physical models. The findings prompt a re-framing of how experimental certainty and repeatability are defined in practice. Conducting repeated trials may mitigate the confounding effects of variability, but the optimal number of trials remains unclear and is likely case-specific. We suggest that future research incorporate repetitions and explore probabilistic approaches to presenting results. Embracing and quantifying this variability can enhance the robustness of physical models and provide deeper insights into channel dynamics.