Tuning an Analysis and Incremental Analysis Updating Assimilation for an Efficient High Resolution Forecast System

For many years the Center for Analysis and Prediction of Storms has been developing and producing real-time high resolution analyses, nowcasts and forecasts. Currently CAPS is supporting the Collaborative Adaptive Sensing of the Atmosphere (CASA) Dallas/Ft Worth Urban Testbed with analyses every 5 minutes and 2-hour forecasts generated every 15 minutes when precipitation is occurring or is expected. These are run on a few hundred Xeon cores. Because of the low-latency demands of the system, we run a cycled Incremental Analysis Updating System (IAU) with increments from the ARPS 3DVAR analysis with complex cloud analysis. The system assimilates experimental surface and profiling data and data from several X-band radars as well as the standard surface and NEXRAD radar data. This work describes recent progress in tuning the analysis and forecast system to improve nowcasts and short term forecasts of ongoing convective storms and introduces variable-dependent timing incremental analysis updating. One element in the analysis product used in these systems is the analysis of hydrometeors, including clouds and precipitation-sized rain, snow, ice, and hail/graupel. The hydrometeor mass assignment has recently been customized for several microphysics schemes. It now uses direct or iterative inversion of radar reflectivity equations for the Lin 5-class Ice microphysics scheme, the WRF single-moment 6-class (WSM6) scheme, Thompson scheme, and Milbrandt and Yau (MY) single-moment bulk microphysics scheme, MY double-moment bulk microphysics scheme, and MY triple-moment bulk microphysics scheme. The standard IAU method applies the increments for all variables with the same distribution of the increments with time. The CAPS systems use a triangular weighting with a greater portion of the increments applied in the middle of a 10-minute time window. It is hypothesized that the retention of hydrometeors aloft would be aided if the wind and latent heat adjustments would lead the introduction of hydrometeors. Variable-dependent timing (VDT) in IAU has been devised to accomplish this. The design of VDT in IAU is presented and tested in cases of ongoing convection in the CASA D/FW Urban Testbed. Figure: Variable-Dependent Timing distribution of increments in Incremental Analysis Updating for temperature, moisture and wind variables (red) and hydrometeor variables (blue).