4.3 Observing System Simulation Experiments Assessing the Potential Impact of Nation-Scale Mesoscale Surface Network on Short-Range Weather Forecasting

Monday, 29 January 2024: 5:00 PM
Key 9 (Hilton Baltimore Inner Harbor)
Chengsi Liu, Center for Analysis and Prediction of Storms, Univ. of Oklahoma, Norman, OK; and T. Sun, M. Xue, and B. Moore III

The skill of numerical weather prediction (NWP), especially for short-range convective-scale NWP, can be substantially improved through assimilation of observations from mesoscale and convective-scale observing platforms. The mesoscale surface observation networks (i.e., mesonets) provide data at higher spatial and temporal resolutions than conventional observing systems (e.g., US ASOS and AWOS), and are routinely used by forecasters in real time to monitor the evolution of rapidly evolving surface mesoscale features that are important in monitoring and forecasting convection initialization (CI). However, only a few states have dense surface mesoscale observing networks, including those of Oklahoma, Kentucky, New York and West Texas. The very limited spatial coverage of these observing systems limits the impact of such data on national-scale weather forecasting, especially for longer forecast ranges beyond several hours.

In this study, we perform Observing System Simulation Experiments (OSSEs) using the JEDI data assimilation (DA) system and the limited area model version of the FV3 (FV3-LAM) to evaluate the potential impact of gap-filling surface station observations on short-range NWP at the convection-allowing resolution. Specifically, nature runs initialized from GFS analysis is performed using WRF model with microphysics different from the assimilation run. By applying observation operators to the nature run fields, we generated simulated observations from existing conventional observations (e.g., sounding and surface observations) and planned national gap-filling mesonet with mean spacing of ~30 km. Simulated conventional observations alone or together with gap-filling Mesonet are assimilated using JEDI LETKF coupled with FV3-LAM.

A series of sensitivity experiments for a single case are first carried out to determine the optimal localization radii, station density and DA frequency. The results show that using a horizontal localization radius of 300 km can better analyze both the large-scale environments and convective-scale storms, and a vertical localization scale of 0.3 scalar height is able to properly transfer the surface information to the upper levels. The sensitivity experiments of Mesonet density (30, 90, 150 km) show that assimilating observations with higher density can provide more fine-grained mesoscale details and improve CI forecast. Additionally assimilating gap-filling Mesonet data in the final hour with 15-minute interval further improves the prediction of convections as well as reduces forecast errors of both surface and upper-level variables. However, the 5-minute DA interval in the final hour degrades the forecast skills possibly due to the imbalance introduced by high-frequency DA.

The potential impact of assimilating the gap-filling Mesonet data are further evaluated by carrying out the 1-week analysis and forecast experiments with the above optimal DA configurations. The forecast skill up to about 6 hours can be improved in terms of higher FSS and lower frequency bias for accumulative precipitation and composite reflectivity forecasts. The verification against the truth simulation also shows that the forecast errors of both upper-level and surface variables are reduced. The detailed evaluations will be presented at the conference.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner