Composite Observation Impact on Short-Range Tropical Cyclone Intensity Forecasts in the Navy Global Environmental Model

While the impact of observations on a tropical cyclone forecast can be computed through an observing system experiment, these experiments are extremely costly and cannot feasibly provide detail down to the impact of individual observations. In this study, the impact of assimilated observations on the forecast intensity of a tropical cyclone is estimated using the adjoint of a numerical weather prediction model – the Navy Global Environmental Model (NAVGEM), and the adjoint of a data assimilation system - the NRL Atmospheric Variational Data Assimilation System – Accelerated Representer (NAVDAS-AR), to provide an individual observation-impact estimate for every observation assimilated into the global analysis. Observation-impact is then binned by observed variable, 3-dimensional location relative to the storm center, and forecast lead-time, and then composited on two-dimensional phase-space grids for evaluation.

Observation-impact on 12-48 hour intensity forecasts for ONR TCI cases Hurricane Joaquin (2015) and Hurricane Matthew (2016) are examined, with impacts composited throughout the life-cycle of each storm. Generalized conclusions about the impact of various observation-types are drawn from the evaluation, as well as an investigation into the uniqueness of value of reconnaissance observations relative to non-reconnaissance observations for Matthew.