07/18/2024
By David Coe

The School of Marine Sciences invites you to attend a doctoral dissertation defense by David Coe on “Regional Weather Typing: Daily Circulation-Based Analysis of Seasonal Transitions and Extremes in the Northeast U.S.”

Candidate: David Coe
Degree: Ph.D. in Marine Science and Technology
Defense Date: Wednesday, July 31, 2024
Time: 10:30 a.m.
Location: Olney Hall Room 312
Thesis/Dissertation Title: “Regional Weather Typing: Daily Circulation-Based Analysis of Seasonal Transitions and Extremes in the Northeast U.S.”

Dissertation Committee

  • Chair Mathew Barlow, Ph.D., Professor, Department of Environmental Earth and Atmospheric Sciences, University of Massachusetts Lowell
  • Frank Colby, Ph.D., Professor, Department of Environmental Earth and Atmospheric Sciences, University of Massachusetts Lowell
  • Christopher Skinner, Ph.D., Associate Professor, Department of Environmental Earth and Atmospheric Sciences, University of Massachusetts Lowell
  • Ellen Douglas, Ph.D., Associate Professor, School for the Environment, University of Massachusetts Boston

Abstract
The aim of weather type (WT) analysis is to identify a region’s characteristic weather patterns, which represent distinct types of storms and sensible weather. Here, the weather and climate of the Northeast U.S. is examined through identification of recurrent daily atmospheric circulation patterns, which are used to analyze extreme events, changes in weather over the course of the seasons, and to evaluate climate model performance from a process-oriented perspective. The weather types are analyzed in observational data for the transition seasons of fall and spring, and in climate models for the fall season. A key result of this research is that, for the Northeast US, a small set of distinct weather types can account for key aspects of surface sensible weather, including extreme events, as well as the transitions between warm and cold seasons, and trends in these transitions associated with climate change.

For the fall season, daily 500 hPa heights, 850-hPa winds, and MSLP from 1979-2018 are input into a k-means clustering algorithm, producing a set of seven circulation patterns that are analyzed in terms of structure, frequency of occurrence through the season, typical progressions between patterns, precipitation and temperature characteristics, and relation to teleconnections. Two common progressions between the patterns are identified, one most frequent in September and one most frequent in mid-October–November. This seasonality allows for a daily circulation-based distinction between early and late season, and shift toward a longer period of warm season patterns and a shorter, delayed period of cold season patterns is identified. This analysis shows that the well-known temperature trends are also associated with distinct changes to the structure of daily weather circulation patterns.

For the spring season, a similar analysis is conducted, producing a set of seven circulation patterns that are analyzed in terms of frequency, seasonal evolution, and relationship with extreme events. Based on daily frequency, the circulation patterns show seasonality with three primarily occurring during the beginning and three primarily occurring at the end of the season. Trend analysis shows a similar pattern to the fall season, with cold season circulation patterns occurring less frequently and warm season patterns occurring more frequently during the beginning of the season. Close links are documented between the weather types and extreme precipitation and temperature events.

Finally, the potential of using the patterns as tools to assess climate model performance is explored. The historical simulations from phase 6 of the Coupled Model Intercomparison Project are examined in terms of weather patterns for the fall season in the Northeast. Both k-means and AI image classification techniques are explored to match climate model data to the observed fall season circulation patterns. The AI image classification model proved superior to the k-means method, labelling data with a 91% accuracy while also showing model capability of reproducing 6 of the 7 observed circulation patterns with a similar seasonal daily frequency. Based on this analysis, irrespective of model resolution, the climate model simulations are shown to reasonably reproduce most but not all of the fall weather patterns.