Dry salt lakes form in arid valleys which are dominated by evaporation and where groundwater or rivers flow in, but not out. Despite the extreme conditions that often prevail above ground in such places, the water table remains near the surface, giving rise to interesting fluid dynamics. Brought there by the buried water, dissolved salts accumulate at the surface and form a rigid crust decorated by exotic patterns. Such dry lakes occur naturally around the world, such as Sua pan in Botswana, Salar de Uyuni in Bolivia and the lakes in the region of Death Valley in the USA. They are recognised as geological marvels: for example, Badwater Basin in Death Valley receives over a million visitors per year, and Salar de Uyuni inspired the landscape of planet Crait in Star Wars’ The Last Jedi (see figures below). In arid locales, human activity can also lead to the formation of dry salt lakes. For example, the Los Angeles aqueduct diverts Owens River and allowed the 280 km^2 Owens Lake to dry into a salt-encrusted plain.
When wind blows, the crust erodes and the resulting salty dust gets transported away, contributing to potentially harmful atmospheric dust and to mineral transport to the oceans. The dust produced by Owens Lake, rich in toxic substances like arsenic, has been recognised as a major health hazard to the population and is being watched by the city of Los Angeles. In spite of their environmental impact, the fluid dynamics of dry lakes are not well understood. The formation of crust patterns, including the polygonal ridges, occurs over a timeline of weeks to months and has been described as part of detailed observations of the crust dynamics, but without considering any coupling with the subsurface convection. We will here consider the coupling between the crust and subsurface convection. For this, we will use a model of porous medium convection for the subsurface flow, driven by a vertical through-flow due to evaporation at the surface and by a crust saturated in salt. This model has been recently proposed by Lasser et al. and only studied in two-dimensions, where it has provided excellent predictions of pattern sizes. Our three-dimensional modelling will extend these results to predict full patterns and their dynamics within realistic domains.
This project will be supervised by Cedric Beaume and Steve Tobias from the University of Leeds and Lucas Goehring from Nottingham Trent University.