Why do Stable Boundary Layer Simulations over Land Sometimes “Crash”?

Over the past decades, numerical weather prediction (NWP) and climate models around the world have been utilizing several ad-hoc approaches (e.g., long-tail stability correction functions, friction velocity limiters) to avoid three interrelated modeling issues -- decoupling of boundary layer from underlying land-surface, run-away cooling, and crashing -- in the context of stably stratified flows. In order to explain these NWP and climate modeling issues, a few researchers performed idealized single column and large-eddy simulations. They documented that these idealized simulations crashed (without much warning) when substantial amounts of downward (negative) sensible heat flux are prescribed as lower surface boundary conditions. Since these crashing simulations were usually accompanied by substantial temperature drop, these simulations were identified as runaway cooling events. In this presentation, we demonstrate that the previously reported crashing events in idealized simulations are not related to the runaway cooling problem experienced by the atmospheric models. The crashing and runaway cooling events in these models occur when surface fluxes become negligibly small (very stable condition). In contrast, the root of the crashing problem in the idealized simulations can be traced back to the prescription of un-physically large downward (negative) sensible heat flux.