Identifying the Microphysical Sensitivities of Mesoscale and Synoptic Precipitation Using an Ensemble Framework

Abstract

Through ensemble sensitivity analysis, this dissertation aims to identify the amount of forecast uncertainty that stems from the representation of mixed-phase cloud microphysics within the Weather Research and Forecasting Model (WRF). The first research thrust focuses on how the evolution of ice crystal shape and choice of ice nucleation parameterization in the Adaptive Habit Microphysics Model (AHM) influences the lake-effect storm that occurred during Intensive Operating Period 4 (IOP4) of the Ontario Winter Lake Effect Systems (OWLeS) Field Campaign. This localized snowstorm produced total liquid-equivalent precipitation amounts up to 17.92 mm during a 16-hour time period, providing a natural laboratory to investigate the ice-liquid partitioning within the cloud and various microphysical process rates, as well as the accumulated precipitation magnitude and its associated spatial distribution. Two nucleation parameterizations were implemented, and aerosol data from a size-resolved Advanced Particle Microphysics (APM) model were ingested into the AHM for use in parameterizing ice and cloud condensation nuclei. Simulations allowing ice crystals to grow nonspherically produced 1.6–2.3% greater precipitation while altering the nucleation parameterization changed the type of accumulating hydrometeors. In addition, all simulations were highly sensitive to the domain resolution and the source of initial and boundary conditions.

The second research thrust aims to identify which microphysics processes may lead to the greatest forecast uncertainty in the OWLeS IOP4 lake-effect storm as microphysical processes within mixed-phase convective clouds can have cascading impacts on cloud properties and resultant precipitation. A microphysical ensemble composed of 24 simulations that differ in the microphysics scheme as well as changes in the representation of aerosol and potential ice nuclei concentrations, ice nucleation parameterizations, rain and ice fall speeds, spectral indices, ice habit assumptions, and the number of moments is used for modeling hydrometeors in each adaptive habit model. Each of these changes to microphysics results in varied precipitation types at the surface; 15 members forecast a mixture of snow, ice, and graupel, seven members forecast only snow and ice, and the remaining two members forecast a combination of snow, ice, graupel, and rain. Observations from an optical disdrometer positioned to the south of the core of the lake-effect storm indicate that 92% of the observed particles were snow and ice, 5% were graupel, and 3% were rain and drizzle. Analysis of observations spanning more than a point location, such as polarimetric radar observations and aircraft measurements of liquid water content, and comparisons to the ensemble provides insight into cloud composition and processes leading to the differences at the surface. Ensemble spread is controlled by hydrometeor type differences spurred by processes or parameters (e.g., ice fall speed) that affect graupel mass.

The ensemble approach used to investigate system microphysical sensitivity in the second thrust is expanded in the third research thrust, where a stochastic perturbed parameterization (SPP) is implemented into WRF and the AHM to investigate the impact of microphysics process rate perturbations on high-intensity precipitation affecting New York State (NYS). This SPP methodology is used to investigate the impact of perturbations on the most active microphysics process rates in a synoptic rain storm with a tropical moisture connection that impacted NYS from 29–30 October 2017. These process rates include vapor deposition onto ice (IDEP) and snow (SNOWDEP), accretion of droplets by rain (CRACCR), and the melting of snow (SMELT). Nine tuning experiments with slightly different spatial, temporal, and amplitude autocorrelation parameters were conducted to elicit those most conducive to the production of physically-sound ensemble spread (i.e., standard deviation) when perturbing IDEP. These parameters are used, along with random number seeds, to generate a stochastic pattern that perturbs the process rate in each model grid cell, thereby producing an ensemble including stochastic microphysics uncertainty. These SPP methods are compared to ensembles involving initial and boundary condition (IC/BC) uncertainty, the Stochastic Perturbed Physics Tendency (SPPT), independent SPPT (iSPPT) perturbation methods, and IC/BC ensembles combined with SPPT, iSPPT, and/or SPP. The performance of the four SPP ensembles are verified against NYS Mesonet observations of 2-m temperature, precipitation, and melting level. The impacts of these perturbations affect the precipitation forecast across the state, with the greatest changes in forecast spread residing in the upper-end of the forecast range. Unexpected relationships among process rates are uncovered when perturbations are applied to each process rate, which are explored further in this work through analysis and comparison of process rate spread and frequency. These relationships are also examined in low- and high-end precipitation regions, where perturbations that directly affect cold or warm cloud processes may non-linearly impact surface rainfall. The influence of SPP methods is tied back to thrusts 1 and 2 through the exploration of two lake-effect storms, including OWLeS IOP4 and a storm in December 2017. This analysis indicates that perturbations to ice deposition affect the QPF differently than other perturbations due to the impact this process has on other related in-cloud processes in both cases. As a whole, this dissertation contributes new knowledge about precipitation responses to the representation of microphysics and associated uncertainty in numerical weather prediction models.