Weathering and erosion of sandstone produces unique landforms1,2 such as arches, alcoves, pedestal rocks and pillars. Gravity-induced stresses have been assumed to not play a role in landform preservation3 and to instead increase weathering rates4,5. Here we show that increased stress within a landform as a result of vertical loading reduces weathering and erosion rates, using laboratory experiments and numerical modelling. We find that when a cube of locked sand exposed to weathering and erosion processes is experimentally subjected to a sufficiently low vertical stress, the vertical sides of the cube progressively disintegrate into individual grains. As the cross-sectional area under the loading decreases, the vertical stress increases until a critical value is reached. At this threshold, fabric interlocking of sand grains causes the granular sediment to behave like a strong, rock-like material, and the remaining load-bearing pillar or pedestal landform is resistant to further erosion. Our experiments are able to reproduce other natural shapes including arches, alcoves and multiple pillars when planar discontinuities, such as bedding planes or fractures, are present. Numerical modelling demonstrates that the stress field is modified by discontinuities to make a variety of shapes stable under fabric interlocking, owing to the negative feedback between stress and erosion. We conclude that the stress field is the primary control of the shape evolution of sandstone landforms.
We acknowledge the help of G.T. Carlig and D. Tingey for assistance with sampling and measurements in the USA and J. Valek and J. Bohac for UCS and triaxial measurements. We also thank M. Audy, M. Sluka, V. Cilek and J. Adamovic for providing photographs and V. Erban, L. Palatinus, A.N. Palmer, K. Zak, J. Mls and T. Fischer for valuable comments on this manuscript. The rain simulator was provided by VUMOP, Prague. This research was funded by the Grant Agency of Charles University (GAUK No. 380511), Czech Science Foundation (GA CR No. 13-28040S), the research plan No. RVO 67985831 and MEYS grant LK21303. The research was also supported in part by the Hintze Fund at Brigham Young University, Provo, Utah, USA.