An investigation of the uncertainties related to the solution of horizontal advection in UAM-IV

Conference · Tue Dec 31 04:00:00 EST 1996

OSTI ID:466231

Odman, M T; Ingram, C L ^[1]; Russell, A G ^[2]

---

1. MCNC-North Carolina Supercomputing Center, Research Triangle Park, NC (United States)
2. Carnegie Mellon Univ., Pittsburgh, PA (United States); and others

As air quality models are increasingly used for making regulatory decisions, it becomes more important to understand the uncertainties involve din these models. This study investigates if the uncertainty introduced by the numerical solution of horizontal advection in the Urban Airshed Model (UAM-IV) biases model results in favor of certain control strategies. After an intensive literature survey and preliminary evaluation of advection algorithms, three schemes were selected for incorporation in UAM-IV. These are the algorithms by Bott, Yamartino and the accurate space derivative method as used by Dabdub and Seinfeld. The test cases used for evaluation included shear and chemically reactive flow problems. The selection was based on higher order accuracy and other desirable properties such as mass conservation and computational efficiency. UAM simulations were performed for the August 26-28, 1987 episode in Southern California using the current advection solver and the three schemes above. In addition to the base emissions three emissions reduction scenarios were simulated (50% ROG, 50% NO{sub x}, and 50% ROG and NO{sub x}). The advection schemes often performed similarly and introduced little bias both for the base case and the reduction scenarios. The differences were most pronounced at the peak ozone concentrations and locations. The schemes predicted very similar ozone levels up to the 99th percentile of the range. Smolarkiewicz led to predictions most different from others. The cumulative distribution functions and the regression analysis showed very similar predictions over most of the prediction range and modeling domain. Replacing the current solver is not expected to affect the choice of control strategy but it may affect the amount of controls. We recommend the use of Bott or Yamartino algorithms in UAM.

• OSTI ID: 466231
• Report Number(s): CONF-9606185--
• Country of Publication: United States
• Language: English

Related Subjects

54 ENVIRONMENTAL SCIENCES
ADVECTION
AIR QUALITY
DATA COVARIANCES
MATHEMATICAL MODELS
URBAN AREAS