ACINN Graduate Seminar - SS 2024
2024-05-29 at 12:00 (on-line and on-site)
A path to scale-aware representation of the convective boundary layer in numerical weather prediction
Mirjana Sakradzija
Department of Geography, Ludwig-Maximilians-Universität (LMU), Munich, Germany
Shallow convective clouds pose a major research challenge in weather and climate science. On a global scale, cloud feedbacks remain the largest source of uncertainty in climate projections (Bony and Dufresne, 2005). Locally, shallow clouds introduce complex feedbacks with soil moisture and vegetation modulating the moisture and energy budgets of the lower atmosphere. Despite several decades-long research efforts, the representation of convective clouds in numerical models remains inadequate. With a decrease in horizontal grid spacing of the models from tens of km to about 1-3 km, the biases in the representation of low-level clouds have evolved from “too few-too bright” low-level clouds (Nam et al., 2012) to either “too low and vastly overestimated” if no shallow-convective parameterization is used, or “too high and too sparse” if a shallow-convective parameterization is inherited and adapted from coarser-resolution models. Over the past decade, new stochastic approaches have emerged to address the representation of convection and clouds at the grid resolutions of several km. Stochastic approaches represent the uncertainty of convective processes and sampling variability of convective elements (clouds or updrafts) that both increase as the model resolution is refined.
I will present the work done at the Hans-Ertel Centre for Weather Research (Deutscher Wetterdienst) on the development of a stochastic shallow convection scheme in the Icosahedral Nonhydrostatic model (ICON) for use at km-scale resolutions (Sakradzija et al., 2015). The scheme is based on the theory of convective statistical ensembles of Craig and Cohen (2006) previously defined for deep convection. The key building block of the scheme is the statistical distribution of the convective mass flux p(m) derived using Large-Eddy Simulations (LES) and spatiotemporal tracking of individual cloud elements. p(m) is a long-tailed distribution with a shape that is highly influenced by the surface Bowen ratio (BR). This link to BR is used in ICON to relate the statistical cloud representation to the land surface conditions. The cloud scheme is applied in several studies using ICON-NWP at resolutions from 1 to 10 km where it influences the cloud cover, spatial patterns of convective organization, and large-scale precipitation patterns, among others. Further developments of the scheme will consider the effects of complex terrain and transitional ABL regimes on p(m), coupling of the stochastic module with a higher-order turbulence closure, and further evaluation and applications at NWP and climate scales. Our project SCALABLe at HErZ is dedicated to these tasks in achieving a scale-aware representation of the ABL and investigating its impact on explicit deep convection.
The content of this webpage is the author’s opinion and the author’s intellectual property. ACINN has the author’s consent to make this information available on the webpage for the announcement of the seminar presentation and in the seminar archive.