Targeting Coordinated Unmanned Aircraft Observations in Hurricanes

This talk will discuss adaptive sampling with unmanned aircraft systems (UAS), which may prove invaluable for collecting targeted low-altitude observations in tropical cyclones (TCs). A major obstacle to controlling aircraft formations in TCs is the strong flow speed, which is often greater than the vehicle speed relative to the flow. The presence of such a flowfield may inhibit a vehicle from making forward progress relative to a ground-fixed reference frame, thus making some trajectories infeasible. In the context of a self-propelled particle model, we generate Lyapunov-based decentralized control algorithms for trajectories that are deemed promising for hurricane sampling. In particular, we accommodate three basic trajectory shapes: circles, folia, and spirographs. Folium trajectories can closely emulate currently operational flight paths, such as the “figure-4” used by the NOAA P-3, while spirograph trajectories provide near-uniform coverage throughout the TC. The spirographs can be manipulated to concentrate higher sampling densities in specific azimuthally symmetric regions of the vortex.

In observing system simulation experiments (OSSEs), we collect UAS observations from the aforementioned trajectories and assimilate them with an ensemble square root filter (EnSRF). We search the trajectory parameter space either exhaustively or stochastically to identify trajectories that lead to the most skillful forecasts. More advanced OSSEs involve live simulations known as “virtual pilot experiments”. Trajectories are chosen based on ensemble diagnostics, and are re-optimized in situ. Since the relative forecast skill improvements of different trajectories aren't known ahead of time, the virtual pilot experiments more realistically assess the potential benefit of a UAS reconnaissance program.

Ongoing work involves extending the results to include multiple vehicles, time-invariant vortices, and azimuthally asymmetric vortices. We will eventually move beyond idealized vortex models, applying our adaptive sampling techniques to full 3-D Weather Research and Forecasting (WRF) simulations.