Dynamics and Predictability of Tropical Cyclogenesis during PREDICT Revealed from Ensemble Analyses and Forecasts with coupled variational-EnKF Data Assimilation Systems

The primary objective of the recently completed NSF-sponsored field experiment Pre-Depression Investigation of Cloud-systems in the Tropics (PREDICT) is to examine the dynamics and predictability of tropical cyclogenesis. During PREDICT of 2010, the NSF/NCAR G-V aircraft collected flight-level and dropsonde observations of multiple missions for four cases (Fiona, ex-Gaston, Karl and Matthew) for examining the environment and mesoscale structure of developing and non-developing pre-depression disturbances. To maximize the benefits of these unique airborne observations in understanding tropical cyclogenesis, two hybrid data assimilation systems (E3DVar and E4DVar) that couple 3D- and 4D-variational methods (3DVar and 4DVar) with the ensemble Kalman filter (EnKF) are applied to construct dynamically consistent 4-dimensional analysis for these tropical disturbances. Our focus is Hurricane Karl (2010) during the pre-genesis stages of the storm's development. The hybrid systems incorporate ensemble-based background error covariance with variational minimization to provide initial conditions for the Weather Research and Forecasting model at convection-permitting resolution. In particular, the E4DVar method benefits from the flow-dependent forecast uncertainty estimation and probabilistic analysis provided from the Ensemble Kalman filter, in addition to four-dimensional trajectory fitting of observations. Flight-level and dropsonde data collected during PREDICT provide considerable coverage of the disturbance from which Karl emerged; more importantly, regions within the tropical wave that are suspected to be important for the eventual cyclogenesis were targeted. This study evaluates the performance of hybrid data assimilation systems for the initialization of convective-scale weather features, and explores the predictability of Karl's development through the ensemble sensitivity analyses and data denial experiments.