Abstract As noted in [Oke], meteorological weather stations located in urban areas often yield unusable and unreliable data due to their proximity to buildings and other obstructions. Direct import of this data into Numerical Weather Prediction (NWP) and Atmospheric Transport and Dispersion (ATD) as well as Urban Area Transport and Dispersion (UATD) models is problematic since the individual site readings represent highly localized behavior, and often do not characterize the neighborhood and grid scale meteorological conditions needed by the models. Due to the localized flow and turbulence characteristics caused by the unique obstructions near each intra-urban site location, established siting procedures [WMO, 1983.a,b] cannot be followed [Oke]. These issues have not stopped the propagation of many intra-urban sited weather stations, as evidenced by DCNet [http://dcnet.atdd.noaa.gov/] and even WeatherBug [http://www.WeatherBug.com/]. Recent field tests, such as the Urban 2000 [Allwine, et al., 2002], Joint Urban 2003 [Allwine et al., 2004] and the Department of Homeland Security, Urban Dispersion Program (DHS UDP) sponsored tests in New York City (MSG05 and UDP Midtown) [UDP] and [Hanna et al., 2006], have taken considerable data deep in the urban street canyons.

In this paper we demonstrate and assess techniques based upon using high-resolution Computational Fluid Dynamics (CFD) model output to provide a means to calibrate, correlate and optimize intra-urban weather station sites so that their readings may more appropriately represent the prevailing climatological conditions. We base our study upon the meteorological data taken during the Joint Urban 2003 (JU2003) Field Test in Oklahoma City, where a number of wind sensors were located deep in the Central Business District area. The JU2003 data is used in conjunction with high-resolution CFD model output to assess techniques such as;

Sensor Correlations: Correlations between urban-sited sensors and the prevailing conditions are derived using a “best-fit” wind field approach that is found by using non-linear constrained optimization techniques in conjunction with pre-computed CFD model output, stored in a “wind field library”. Correlations between intra-urban met station readings and the conditions at various locations in the urban area are found by using the CFD results, and furthermore, these are correlated with the prevailing conditions. This allows a functional relationship to be found relating “trusted” meteorological stations located upstream of the city with the deep, in-canyon flow fields.

Sensor Placement: The sensitivity of sensor locations relative to building surfaces is addressed, and sensor placement within the urban canopy is also assessed using the high-resolution CFD model output. The effect of sensor placement upon sensor correlation is also assessed by evaluating the sensitivity of the correlations to sensor location for given sensor sites.

Sensor Corrections: Corrections for proximity to building surfaces is found, and is used to better correlate given sensors to the prevailing conditions.

Acknowledgements This work is funded by the Defense Threat Reduction Agency under a Small Business Innovation Research Phase I grant, with Technical Monitor, CDR Stephanie Hamilton.