Abstract:

This research, conducted at NOAA’s National Severe Storms Laboratory (NSSL), developed a framework to close critical data gaps in boundary-layer observations, a region essential for predicting severe weather. Traditional uncrewed aerial system (UAS) flights follow fixed paths, limiting adaptability to evolving atmospheric conditions. To address this, we introduced an unsupervised clustering algorithm trained on radiosonde data from the Integrated Global Radiosonde Archive (IGRA). The algorithm detects structural patterns in vertical temperature profiles and informs adaptive sampling strategies. The framework allows a UAS to first collect a baseline profile, then compare new flight data against clustering results to determine which atmospheric layers are undersampled or highly variable. The UAS can then adjust its flight behavior, spending more time in regions with sparse data rather than distributing measurements uniformly. Implementation combined Python-based preprocessing and clustering workflows with MATLAB and ArduPilot simulations, integrating the approach into CopterSonde, a boundary-layer UAS platform already used in NOAA OAR laboratories. Early tests show the feasibility of real-time adaptive sampling. This work paves the way for UAS operations that actively reduce data gaps, sharpen the resolution of boundary-layer measurements, and strengthen NOAA’s forecasting capabilities in support of the Weather-Ready Nation initiative. The results are from the NOAA EPP/MSI CSCNERTO graduate internship project conducted with NOAA mentor Dr. Elizabeth Smith of the Oceanic and Atmospheric Research, National Severe Storms Laboratory. The NERTO aligns with NOAA CSC CESSRST-II’s mission to advance earth system science, remote sensing, and data-driven environmental solutions, in support of NOAA’s goals of a Weather-Ready Nation, Resilient Coastal Communities, and Climate Adaptation. The NERTO Exploring sampling approaches for NSSL’s UAS applications also deepened the intern’s understanding of NOAA operational workflows, scientific communication, and collaborative research environments, while strengthening professional skills such as technical reporting, cross-disciplinary teamwork, and real-time presentation of scientific results.