Practical Predictability of Tropical Cyclones: Hurricanes Karl (2010), Sandy (2012), and Edouard (2014)

Open Access
- Author:
- Melhauser, Christopher Lee
- Graduate Program:
- Meteorology
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 11, 2016
- Committee Members:
- Fuqing Zhang, Dissertation Advisor/Co-Advisor
Fuqing Zhang, Committee Chair/Co-Chair
Eugene Edmund Clothiaux, Committee Member
Jerry Y Harrington, Committee Member
Steven J Greybush, Committee Member
Zhibiao Zhao, Committee Member - Keywords:
- tropical cyclones
predictability
numerical weather prediction
data assimilation
airborne radar
diurnal cycle - Abstract:
- Various aspects of the practical predictability – the limit on atmospheric prediction using the current optimal analysis procedures to derive the initial state and the best available atmospheric model forecast – of tropical cyclones (TC) are examined using Hurricane Karl (2010), Hurricane Sandy (2012), and Hurricane Edouard (2014) as case studies. The practical predictability of TCs is limited by uncertainties in the forecast model and initial conditions. The uncertainties include the adequacy of observations (i.e. accuracy, spatial and temporal coverage, and usability), data assimilation procedures, and deficiencies in the forecast models. This dissertation is partitioned into three parts that explore: (I) model uncertainties by examining the sensitivity of a model pre-genesis environment of a developing TC to the radiation parameterization, (II) the use of wind retrievals for TC prediction by creating a simplified wind retrieval algorithm using tail-Doppler radar measurements of TCs and generating low error wind field estimates for the purpose of TC data assimilation, and (III) model uncertainties by examining the impact of model dynamic core and physics configurations on the ensemble mean and spread using identical initial conditions to initialize three TC-tuned regional models. In Part I, if was found that the pre-genesis environmental stability and intensity of deep moist convection associated with a TC can be considerably modulated by the diurnal extremes in radiation. In Part II, it was found that assimilating u-wind and v-wind from a new simplified coplane retrieval algorithm can impact track and intensity forecasts differently compared to assimilating direct radial velocity even though similar information is being assimilated. The computational efficiency and real-time low error wind field retrieval of the new simplified coplane retrieval algorithm make this a competitive method for observation generation for TC prediction. Finally, the findings in Part III suggest that model physical parameterizations are a dominant source of model error and can drastically impact ensemble forecast statistics. It is hypothesized that for the two case studies presented, an ensemble can be constructed using a single model core that has similar forecast error characteristics as an ensemble constructed from a multiple model cores.