Quantifying the impact of assimilating sensitive observations on heavy precipitation forecast with ensemble-based forecast sensitivity to observations technique

Abstract

In this study, the Ensemble-based Forecast Sensitivity to Observations (EFSO) technique was used to examine the impact of assimilated sensitive observations on the forecast of heavy rainfall through three data assimilation strategies, i.e., running without observation assimilation (CNTL); with assimilation of all observations (CNTL_DA); with assimilation of sensitive observations (SEN_DA). The Infrared Atmospheric Sounder Interferometer presented the largest positive impact on the forecast, followed by the Advanced Microwave Sounding Unit A and surface land reports. The top 10% of observations contributing the largest positive impact on the forecast were located in the upstream ocean of the low-level winds associated with precipitation maximum. Compared to ensemble CNTL and CNTL_DA experiments, ensemble SEN_DA exhibited better precipitation probability, in which the representative member in ensemble SEN_DA successfully predicted the precipitation magnitude and location. Based upon the representative experiments of the ensemble CNTL, CNTL_DA, and SEN_DA, synoptic conditions were examined to understand how the SEN_DA improves the forecast of heavy precipitation. The three representative experiments exhibited similar characteristics and variations in large-scale circulation, horizontal water vapor transport, and wind fields. However, compared to the representative experiments of CNTL and CNTL_DA, the advection and barotropic components of potential divergence contributed to the increase and persistence of convective instability throughout heavy rainfall in SEN_DA. Heavy precipitation was closely associated with the intensity, duration, and location of the maximum convective instability. Furthermore, the release of convective instability depended on the mesoscale ascent, which determined the maximum and location of the precipitation. These results help to understand the optimization of observational design and assimilation strategy.