Systematic biases in WRF dependent on prevailing land-atmosphere coupling regime

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 5 January 2015
Phoenix Convention Center - West and North Buildings
Craig R. Ferguson, Atmospheric Sciences Research Center, University at Albany, SUNY, Albany, NY; and H. J. Song

Previous research using satellite remote sensing and atmospheric reanalyses applied the joint probability space of convective triggering potential (CTP) low-level humidity index (HI) surface soil moisture (SM) to classify days into one of four land-atmosphere coupling regimes: wet advantage, dry advantage, transition, and atmospherically (remotely)-controlled. Wet- and dry-advantage regimes correspond with periods when wet and dry surface soil moisture anomalies, respectively, favor convective triggering. During transition periods the land signal is mixed; the land signal is reduced to noise in atmospherically-controlled cases. Therefore, if ever we could expect an improvement in forecasts due to land initialization or assimilation of land variables it would be during wet- and dry-advantage regimes. Our hypothesis is that default WRF-Noah is systematically biased in each of these regimes in a way that is distinctly different, and further, that these differences may be exploited to improve coupled models and land data assimilation best practices. This study is focused on the semi-arid U.S. Southern Great Plains (SGP), an area known for potentially strong warm season land-atmosphere coupling. We use (5) persistent wet and (5) persistent dry-advantage regime events from the recent historical record to establish default WRF-Noah error correlation and covariance analogues for each regime. These analogues comprise a coupling chain of forecast error correspondent with the soil moisture-precipitation feedback pathway. Therefore, they afford insights into the aspects/ vertical model layers where the greatest gains in model forecast skill can be expected from land initialization and/or assimilation. Indeed, we perform runs for the identical wet- and dry-advantage periods with land initialization from the North American Land Data Assimilation System Phase 2 (NLDAS-2)-Noah and analyze forecast improvements relative to the error correlation and covariance analogues. We illustrate the contrasting shortcomings and potential gains in forecast skill for both the wet- and dry-advantage regimes. Initial conditions and boundary forcing is taken from the North American Regional Reanalysis (NARR). The forecast period is from 0-36hrs.