Quantifying UAS Observation Error Variance Used in Data Assimilation Systems and Its Impact on Predictive Skill

Abstract

Observation error determines the weights of the observations and background state used in data assimilation to generate analyses. Quantifying observation error is critical for the optimal assimilation of observational data sets. Uncrewed Aircraft System (UAS) observations have shown potential benefits in filling observational gaps in the lower atmosphere; however, characterization of their error characteristics has been limited. To optimize the use of UAS observations in numerical weather prediction, UAS observation error is estimated based on the 3-cornered hat diagnostic approach which uses three independent estimates of the atmospheric state. This approach is applied to data from the 2018 Lower Atmospheric Profiling Studies at Elevation-a Remotely-piloted Aircraft Team Experiment field campaign using collocated UAS and rawinsonde observations along with output from a set of convection-permitting model simulations. The estimated observation error values for UAS temperature, wind, and relative humidity measurements were found to be only weakly dependent on height AGL with mean values equal to 0.5°C, 0.8 m s⁻¹, and 3%, respectively. Only the newly estimated observation error for temperature differed from that previously used to assimilate commercial aircraft observations into global models (1.0°C). However, using this reduced temperature observation error produced more accurate mesoscale analyses and forecasts of both terrain-driven flows and convection initiation generated by colliding outflow boundaries within the San Luis Valley of Colorado.