This talk will discuss a framework for spatially and temporally continuous adaptive sampling in tropical cyclones. We have implemented an ensemble square root filter (EnSRF) to enable serial assimilation of synthetic unmanned aircraft observations. The initial objective is to identify optimal sampling strategies for state estimation of a time-invariant Rankine vortex. We generate a set of minimally parameterized candidate aircraft trajectories, consisting of fundamental patterns such as lines, circles, spirographs, and folia. Vehicle control laws to achieve these trajectories are detailed in a companion presentation. We seek the candidate trajectories that most efficiently reduce ensemble spread and various analysis error metrics. To accomplish this, the parameter space is searched either exhaustively or stochastically, depending on its dimensionality. Results are further extended to include multiple vehicles, time-invariant vortices, and azimuthally asymmetric vortices.

The talk will also present preliminary results from a series of observing system simulation experiments (OSSEs). Each OSSE covers the entire flight duration of a low-altitude, high-endurance UAV, encompassing frequent assimilation cycles. At the beginning of an OSSE, trajectories are selected using the optimization routine described previously. The trajectories are flown through the nature vortex using a virtual aircraft model; observations are collected and assimilated immediately into the EnSRF. The latest state estimates are used to assess whether the trajectories require in-situ re-optimization, establishing a feedback loop between the EnSRF and the aircraft. Forecasts initialized at each assimilation time are probabilistically verified, quantifying the minimum achievable baseline uncertainty as a function of both lead time and the number of aircraft.

Ongoing work involves migrating the OSSEs from Rankine vortices to high-resolution, three-dimensional numerical simulations with the Weather Research and Forecasting (WRF) model.