Observing System Simulation Experiments to Investigate Impacts of Deep-Tropospheric UAS Profiles on NWP Forecasts
Location
Mori Hosseini Student Union & Events Center (Rooms - TBD)
Keywords
UAS, OSSE
Presenter Abstract
Radiosondes and commercial aircraft are the primary sources of in-situ meteorological data in the upper troposphere. While robust, these systems leave major coverage gaps which can lead to forecast errors in both global and regional numerical weather prediction. Assimilating data collected by Weather-observing uncrewed aircraft systems (WxUAS) have shown potential to yield large improvements in NWP forecast accuracy via profiles up to 2km AGL (Murdzek et al. 2026), but the impact of assimilating WxUAS profiles through the deep troposphere remains unknown. Before deploying WxUAS at scale, it is important to quantify the impact that simulated WxUAS observations have on NWP relative to current observations and compare the relative benefits of different WxUAS network configurations; additionally, it is necessary to quantify the capabilities required of WxUAS platforms to achieve these benefits.
To address these questions, we designed an observing system simulation experiment (OSSE) with simulated WxUAS networks performing vertical profiles of the deep troposphere (up to 15km AGL) over the CONUS. This OSSE framework comprises a 1-km WRF simulation over 2 week-long periods in winter and spring, software for creating realistic conventional and synthetic WxUAS observations, and a HRRR-like forecast system. Data-denial experiments are performed replacing synthetic radiosondes with an idealized grid of WxUAS sites combined with assimilated data from commercial aircraft, surface stations, and GPS-derived precipitable water. Various network configurations of WxUAS are tested with a baseline experiment launching WxUAS at a similar network density (~70 sites) and launch frequency (every 12 hours) to radiosondes. Additional experiments address the impacts of adding more sites vs. profiling more frequently, changing profile depth, and restricting WxUAS at different wind and icing thresholds. Resulting verification using root-mean-square errors (RMSEs) against 3D distributions of temperature, relative humidity, winds, and convective parameters (spring only) will be shown. Initial findings suggest that a 71-site deployment of WxUAS reduces forecast RMSEs by up to 5% over forecasts assimilating radiosondes when the WxUAS are launched every 2 hours. Only fractional benefits over forecasts with radiosondes are realized when reducing all profile depths or limiting flight heights by applying limits due to wind/icing.
Presentations
Presented in Session 10: Data Modeling
Observing System Simulation Experiments to Investigate Impacts of Deep-Tropospheric UAS Profiles on NWP Forecasts
Mori Hosseini Student Union & Events Center (Rooms - TBD)
Radiosondes and commercial aircraft are the primary sources of in-situ meteorological data in the upper troposphere. While robust, these systems leave major coverage gaps which can lead to forecast errors in both global and regional numerical weather prediction. Assimilating data collected by Weather-observing uncrewed aircraft systems (WxUAS) have shown potential to yield large improvements in NWP forecast accuracy via profiles up to 2km AGL (Murdzek et al. 2026), but the impact of assimilating WxUAS profiles through the deep troposphere remains unknown. Before deploying WxUAS at scale, it is important to quantify the impact that simulated WxUAS observations have on NWP relative to current observations and compare the relative benefits of different WxUAS network configurations; additionally, it is necessary to quantify the capabilities required of WxUAS platforms to achieve these benefits.
To address these questions, we designed an observing system simulation experiment (OSSE) with simulated WxUAS networks performing vertical profiles of the deep troposphere (up to 15km AGL) over the CONUS. This OSSE framework comprises a 1-km WRF simulation over 2 week-long periods in winter and spring, software for creating realistic conventional and synthetic WxUAS observations, and a HRRR-like forecast system. Data-denial experiments are performed replacing synthetic radiosondes with an idealized grid of WxUAS sites combined with assimilated data from commercial aircraft, surface stations, and GPS-derived precipitable water. Various network configurations of WxUAS are tested with a baseline experiment launching WxUAS at a similar network density (~70 sites) and launch frequency (every 12 hours) to radiosondes. Additional experiments address the impacts of adding more sites vs. profiling more frequently, changing profile depth, and restricting WxUAS at different wind and icing thresholds. Resulting verification using root-mean-square errors (RMSEs) against 3D distributions of temperature, relative humidity, winds, and convective parameters (spring only) will be shown. Initial findings suggest that a 71-site deployment of WxUAS reduces forecast RMSEs by up to 5% over forecasts assimilating radiosondes when the WxUAS are launched every 2 hours. Only fractional benefits over forecasts with radiosondes are realized when reducing all profile depths or limiting flight heights by applying limits due to wind/icing.