Impacts of temporal and spatial resolutions of model initialization data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 18 January 2010
Exhibit Hall B2 (GWCC)
T.J. Malone, Western Kentucky University, Bowling Green, KY; and R. D. Leeper, R. Mahmood, and A. I. Quintanar

There are a number of available datasets for model initialization at various temporal and spatial resolutions. For instance, NCEP's Final Analysis Data (FNL) at 1x1 (110x110 km) is only available every six hours. However, NCEP's North American Regional Reanalysis (NARR) has a higher resolution at 32x32 km and available every 3 hours. It is often perceived that initial conditions at higher resolution add value to the quality of model simulations through better resolved localized atmospheric features. In order to assess the added value of higher temporal and spatial resolution on model results, a series of 24-hour simulations were conducted using FNL and NARR datasets. Temporally, lateral conditions were forced every six (FNL and NARR) and three hours (NARR). In addition, model domain resolution was varied between 4 and 12 km to study the significance of horizontal resolution on simulated results with a total of six simulations. These results were compared with observational data from the Kentucky Mesonet. Preliminary r2 results suggest the model was a better predictor of two-meter temperature than relative humidity compared to observations. However, increased spatial resolution from FNL to NARR improved simulated relative humidity compared to observations at 6 out of the 10 sites. Furthermore, simulations laterally forced every 3 hours did not outperform the 6 hour simulation. In some instances, increased temporal resolution diminished simulated results compared to observations for both temperature and atmospheric moisture.