Three notable new features of the CAPS efforts in support of HMT this year included 1) implementation of a localized probability matched mean ensemble averaging technique, 2) development of a new spatially-aligned mean algorithm, and 3) preliminary testing the FV3 forecast model with Thompson microphysics.
The localized probability matched mean (LPM) algorithm is designed to confer many of the benefits of the global probability matched mean (PM) while better capturing local convective structures within the forecast ensemble. Rather than performing the PM calculations over the entire CONUS domain, the LPM divides the domain into overlapping patches and the PM algorithm is applied to local patches.
Common differences noted among various ensemble forecast members are differences in storm initiation timing, location and propagation phase speed. Combined these differences lead to offsets in location of convective features among the ensemble members. Averaging members with such offsets creates blurring and smoothing in the resultant ensemble mean products. The Spatially-Aligned Mean (SAM) is designed to reduce these affects by first spatially aligning the 2D products to a common location. The method for CAPS SAM is described, and testing is done with recent CAPS ensemble output.
CAPS implemented the Thompson microphysics scheme in the FV3 model because it is expected to perform better at the convection-allowing 3-km grid scale than the original GFDL microphysics developed for global scale. The implementation of the Thompson scheme available for the 2017 HWT and HMT FFaIR is described and preliminary evaluation of the FV3-Thompson forecast results are presented.
Figure Caption: Comparison of ensemble mean results from 19 May 2017 (HWT Ensemble, used for development and testing). Upper Left: Observed MRMS Composite Reflectivity; Lower Left: Standard Probability Matched Mean; Lower Right: Localized Probability Matched Mean.