The influence of misrepresenting the nocturnal boundary layer on idealized convection in large-eddy simulation

Numerical weather prediction (NWP) is rapidly approaching the convective grey zone, and with ever increasing computational resources, employing large-eddy simulation (LES) to avoid grey zone issues becomes conceivable. Also, with grid spacings of O(100m), this allows for an explicit representation of (shallow) convection; one of the major uncertainties in NWP and climate models. However, a grid spacing of O(100m) is insufficient to resolve many other processes like small-scale shear driven turbulence in the stable nocturnal boundary layer (NBL), potentially introducing biases in its development, and the skill of LES to faithfully represent the diurnal cycle of convection.

We systematically study the influence of misrepresenting the nocturnal boundary layer on daytime convection in low-resolution LES. Guided by measurement data from meteorological sites across Northern Europe, the typical summertime NBL conditions are first characterized. Subsequently, these conditions are used to design idealized numerical experiments of diurnal cycles of dry convection at varying meteorological conditions, and (horizontal) grid spacing ranging from 3.125 m to 100 m. Using the UCLA-LES code with a Smagorinsky sub-grid scheme and simplified version of the ECMWF land-surface model, we study the ability of (low resolution) LES to represent the nocturnal conditions, and influence of potential biases in the NBL on the consecutive day of convection.

Our findings indicate that the use of insufficient resolution introduces significant biases in the NBL, but that the influence on daytime convection is limited, with maximum biases in the afternoon mixed-layer depth and temperature of approximately 100 m and 0.5 K, respectively. In terms of the mixed-layer top relative humidity, these biases partially compensate, resulting in relative humidity biases in the order of 2-3%.