Date of Award

8-1-2021

Language

English

Document Type

Master's Thesis

Degree Name

Master of Science (MS)

College/School/Department

Department of Atmospheric and Environmental Sciences

Content Description

1 online resource (vii, 171 pages) : color illustrations, color maps.

Dissertation/Thesis Chair

Justin Minder

Keywords

Freezing rain, Precipitation type, Weather forecasting, Winter weather, Precipitation forecasting, Boundary layer (Meteorology), Freezing precipitation, Microphysics

Subject Categories

Atmospheric Sciences

Abstract

Predictability challenges are heightened in winter weather forecasting when the environment for high-impact weather is marginal or varies over short distances. High-resolution numerical weather prediction (NWP), such as the High-Resolution Rapid Refresh (HRRR), and ensemble forecast systems, such as the Global Ensemble Forecast System (GEFS), are useful for constraining forecasts. However, their use can be challenging in marginal, near-freezing, situations when precipitation type is uncertain. Uncertainties in planetary boundary layer (PBL) and microphysics (MP) parameterizations and subtle synoptic-scale model errors brought on by differences in initial and lateral boundary conditions (IC/BCs) complicate the p-type forecast. Further, complex terrain, such as that of the Mohawk and Hudson Valleys of Eastern NY, induce significant changes in temperature, wind, and p-type over short distances. In combination, these factors can exacerbate uncertainty and lower forecast confidence. Two recent winter weather events are intensively studied in this project. The first occurred on 6-7 February 2020, when a multi-phase event impacted Eastern NY. The greatest period of impacts, between 12 and 21 UTC 07 February, featured a strongly forced deformation band bring rain and freezing rain, switching to a brief period of heavy snow. 6 to 12 inches of snow fell in the Mohawk and Upper Hudson Valleys, and 0.25 to 0.5 inch of ice accreted – higher than operationally forecasted – north of Albany into the southeastern Adirondacks. The second event occurred between 18 UTC 23 March 2020 and 03 UTC 24 March 2020. Precipitation was forced by a broad area of warm-air advection aloft as a surface low pressure developed along the mid-Atlantic coast. Snow fell in the Hudson and Mohawk Valleys and surrounding mountains during this time frame, and the heaviest amounts of 6 to 10 inches fell in the upper Hudson Valley and eastern Adirondacks while the Capital Region and eastern Catskills received 4 to 7 inches. Snowfall was consistently under-forecast in much of Eastern NY. This study analyzes the Weather Research and Forecasting (WRF) model’s performance in simulating temperature, moisture, wind, and precipitation during the event. The innermost domain, centered over the Northeast US, uses a HRRR-like, 3-km grid. In the PBL and MP ensembles of experiments, PBL and MP parameterizations are altered, respectively, to determine how uncertainty in PBL and MP processes contribute to forecast uncertainty. In the IC/BC ensemble, experiments are conducted with IC/BCs supplied by each of the 21 GEFS members. In both sets of experiments, we evaluated the ensemble’s performance against ASOS and NYS Mesonet observations, mPING reports, and ERA-5 reanalysis datasets. By analyzing variability across the ensemble, we diagnosed the mechanisms, such as temperature advection by terrain-channeled flow, whereby minor synoptic uncertainties translated into considerable differences in surface p-type. Variability in the PBL and MP ensembles was smaller than expected, but both ensembles produced vastly different outcomes than what was observed due to a warm bias. This warm bias was caused by subtle shifts in synoptic features, such as the track of the surface low pressure center, leading to a reversal in the modeled valley-level wind direction and thereby a reversal in the sign of the temperature advection. The IC/BC ensemble featured more variability and less of a warm bias than the physics ensembles, and synoptic-scale features were more adequately represented. Differences between the outcomes of the two ensembles can be attributed to differences in the representation of synoptic features and the associated effects on mesoscale circulations, as well as the differences in initialization datasets between the two sets of experiments.

Download

Share

COinS