"Towards Improving the Detection of North Atlantic Tropical Cyclogenesis (TCG)"

Open Access
Waters, Jeffrey Joseph
Graduate Program:
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 05, 2011
Committee Members:
  • Jenni Evans, Thesis Advisor
  • Chris Eliot Forest, Thesis Advisor
  • logistic regression
  • principal component analysis
  • sub-seasonal variability
  • GCM
  • general circulation model
  • global climate model
  • tropical cyclogenesis
  • tropical cyclone
  • genesis
  • tropical meteorology
The primary focus of this study is to identify a combination of independent thermodynamic and dynamic parameters that accurately detect realistic environments conducive for tropical cyclogenesis (TCG) based on their anomalous sub-seasonal variability within atmospheric general circulation model (GCM) simulations. In particular, we assess the capabilities of an atmospheric GCM for detecting large scale and localized conditions for TCG across the North Atlantic Main Development Region (MDR: 9°-21°N, 20°-80°W) on 10-day and 15-day timescales. We also evaluate sub-seasonal variability, estimated as the standard deviation of daily data, as a predictive tool for TCG likelihood. Using the NCEP/NCAR and ERA-40 reanalysis datasets as a benchmark this study assesses GCM daily output from a 4-member ensemble of the Community Atmospheric Model version 3.1 (CAM3.1) forced with observed monthly sea-surface temperatures over the period June-September 1981-2000. We identify 17 thermodynamic and dynamic TCG variables as potential predictors. From these variables, we create four categories: thermodynamic-convective, dynamic-environmental, dynamic-rotational, and combined thermodynamic-dynamic. Overall, the GCM ensemble exhibits skill at simulating large-scale TCG environments on the seasonal scale and localized TCG environments on the daily scale. To assess sub-seasonal variability of each TCG variable, MDR-averaged 10-day and 15-day standardized deviation anomalies are calculated across the CAM3.1 and ERA-40 datasets. Comparisons of results from a principal component analysis (PCA) applied to CAM3.1 and ERA-40 data indicate that CAM3.1 identifies similar independent predictors for the thermodynamic-convective and dynamic-environmental categories. However, in the dynamic-rotational category, CAM3.1 is unable to capture vertical relative vorticity gradient variability. When the thermodynamic-dynamic category is evaluated, dynamic-rotational terms show larger variance in CAM3.1 compared to other categories while equal levels of variance are seen in ERA-40. To assess predictive skill of TCG occurrences, a logistic regression technique is implemented on the ERA-40 data and corresponding TCG events. Sub-seasonal variability on 15-day timescales performs well as a diagnostic tool for TCG occurrences, yielded high accuracy (68.1-78.8%) and moderate precision (37.5-71.4%) totals, however, threat scores are considerably lower (5.6-42.4%). Such findings indicate the potential for applying this method to CAM3.1 output and corresponding model-simulated TCG occurrences.