Long Duration Stratospheric Balloon Drift and Soundings and Weather Forecasting Impacts in NWP and AIWP

High-resolution forecasting systems and emerging AI-based weather models are placing increasing demands on atmospheric observations, particularly for vertically resolved measurements in data-sparse regions and across the boundary layer–upper troposphere/lower stratosphere (UT/LS) interface. Current observing networks, however, remain limited in their ability to provide persistent, co-located thermodynamic and kinematic measurements with sufficient temporal and vertical resolution.

Here we present a new in-situ observing approach based on long-duration stratospheric balloons. These platforms provide repeated high-resolution vertical profiles (surface to ~14 km) together with multi-day Lagrangian float trajectories in the upper troposphere. This dual capability enables simultaneous sampling of boundary layer structure, free-tropospheric thermodynamics, and upper-level wind evolution, offering a uniquely comprehensive dataset for process studies, model evaluation, and data assimilation.

System performance is demonstrated through coordinated measurement campaigns, including validation against independent aircraft observations and characterization of vertical resolution (10–100 m in the boundary layer, with ~100 levels per profile). The platform provides persistent sampling across the full diurnal cycle and extends coverage into oceanic and other data-sparse regions, addressing key observational gaps for next-generation NWP and AI-based forecasting systems.

To assess the value of these observations, we conduct Observing System Experiments (OSEs) within an ensemble data assimilation framework coupled to an AI-based forecast model (AIFS-ENS). We separately assimilate (1) vertical thermodynamic profiles, (2) Lagrangian wind observations, and (3) the combined dataset. Results show that vertical profiles primarily improve thermodynamic structure and boundary layer representation, while Lagrangian measurements provide strong constraints on upper-level winds and jet positioning. The combined dataset yields complementary benefits.

These results highlight the potential of long-duration balloon systems as a new class of atmospheric observing networks, capable of delivering calibrated, high-resolution measurements for model development, satellite validation, and operational assimilation. They also demonstrate a practical pathway for integrating advanced observational platforms with AI-based prediction systems, supporting the development of observation-adaptive forecasting.