Hierarchical Testing for Improvement of Stochastic and Deterministic Physical Parameterizations within the Unified Forecast System (UFS)

Physical parameterizations are approximate solutions to physical processes occurring in a grid-box and are as such a source of forecast model errors and uncertainty. Advancing physical parameterizations is a key way to improve the skill of numerical weather prediction systems. While stochastic physics schemes are often tuned using ad-hoc methods and informed by expert knowledge, objective methods derived from physical constraints can be used to better inform the development and improvement of schemes. The objective method to inform the development of deterministic and stochastic schemes used for this study follows published methodologies referred to as “coarse-graining”. This method compares the state variables and tendencies in a convection-permitting high-resolution simulation against a lower-resolution parameterized-convection simulation. The high-resolution simulations contain information about the subgrid-scale uncertainty that can be used to objectively inform stochastic parameterizations. A Single Column Model (SCM) incorporates the explicit physics parameterization schemes from global models and is an entry point in the hierarchical testing framework. Forcing a SCM with the coarse-grained data from high-resolution runs removes the problem of the convection-resolving and -permitting models tending to drift away from each other, and enables one to focus on the physical parameterization aspects.

The Model Uncertainty Model Intercomparison Project (MU-MIP) is an international effort to better understand model physics uncertainty, and how to represent it in stochastic physical parameterizations. The Developmental Testbed Center has recently begun contributing to this effort using the Common Community Physics Package (CCPP) SCM. An array of 44,000 SCM simulations over the Indian Ocean were conducted using two NOAA operationally relevant physics suites (Global Forecast System [GFS] v17 prototype 8 and the Rapid Refresh [RAP]). The SCM was forced by coarse-grained 3-km German weather service Icosahedral Nonhydrostatic (ICON) model output, initialized every 3 hours. Preliminary diagnostics and analysis of the systematic differences between the SCM and coarse-grained high-resolution fields, including biases for state variables, will be shown. Probability Density Functions (PDFs) of physics tendencies for temperature, wind components, and humidity produced by the SCM using the GFS and RAP physics suites will be compared with ICON data to obtain an objective estimate of subgrid-scale uncertainty. Finally, ongoing work to force the SCM using coarse-grained high-resolution Unified Forecast System (UFS) forecasts will be discussed.