Executive Summary This report describes the development of algorithms to derive accurate surface roughness and displacement height values from NASA satellite data for both urban and vegetation environments and to evaluate the potential benefit of improved surface data characterization on atmospheric transport and dispersion modeling. One of the key factors limiting the accuracy of near-surface wind profile modeling and dispersion simulations is the lack of location-specific and

more »

... ent surface characteristic values. Currently, land use category look-up tables are commonly used to provide grid-cell estimates of surface characteristics values, even though surface roughness values can vary significantly within the same land-use category. This report documents: i) the derivation of surface characteristic values by Goddard Space Flight Center (GSFC) and Stennis Space Center (SSC) from NASA satellite multispectral imagery, ii) their incorporation into atmospheric meteorological and transport models, and iii) the results of a dispersion modeling study for Oklahoma City and the surrounding region to evaluate the impact of GSFC/SSC roughness on National Atmospheric Release Advisory Center (NARAC) / Interagency Modeling and Atmospheric Assessment Center (IMAAC) emergency response dispersion predictions. The NASA satellite derived surface data evaluation study is performed based on the following steps:  Extension of recently developed and published GSFC/SSC methodologies to generate urban and non-urban surface roughness and displacement height fields covering the continental United States by utilizing NASA Moderate Resolution Imaging Spectroradiometer (MODIS), Landsat, and Shuttle Radar Topography Mission (SRTM) satellite data  Generation of combined urban and non-urban surface characteristics data fields for NARAC/IMAAC dispersion sensitivity simulations over a research domain centered over Oklahoma City  Development of a capability to ingest GSFC/SSC-provided surface roughness and displacement height fields into the NARAC/IMAAC modeling system  Investigation of the sensitivity of atmospheric transport and dispersion predictions to incorporation of GSFC/SSC-derived surface roughness fields by simulating both CONTROL dispersion runs (using default NARAC/IMAAC surface data parameter values) and NASADATA runs (which use both NASA satellite derived surface roughness and displacement height fields)  Formulation of recommendations on future work to further evaluate the potential benefits of using NASA surface roughness products in dispersion modeling decision support tools 2 Conclusions The scientific conclusions resulting from the data evaluation study are summarized below:  A systematic difference in GSFC/SSC roughness lengths for rural land use is observed over the Oklahoma City research domain, with CONTROL values predominantly higher than NASADATA by around 0.05 to 0.1 m for crop and grassland and up to 0.45 m for deciduous forest land use. The typical difference between CONTROL and GSFC/SSC urban roughness lengths is between -0.2 to 0.5 m with CONTROL values predominately larger.  Use of GSFC/SSC surface characteristic values derived from NASA satellite data decreases simulated near surface wind speeds in rural areas by around 0.25 m/s and increases surface layer wind speeds at heights between 15 m and 130 m by 0.25 to 0.75 m/s. Simulated nearsurface wind speeds in urban areas are significantly influenced by using GSFC/SSC surface roughness and displacement height values. CONTROL run urban wind speeds below a height of 10 m are frequently 0.5 to 1.0 m/s less than those generated based on NASADATA surface data. Simulated wind speeds at heights between 10 and 130 m often increase by a few meters per second when using GSFC/SSC-derived surface data.  Different statistical metrics of dispersion prediction skill indicate comparable performance between CONTROL and NASADATA runs. Runs based on CONTROL default surface parameter values provide the best dispersion results based on the standard fraction of predicted values within a factor or 2, 5, and 10 of the observed concentration (FAC2, FAC5, and FAC10) and the geometric mean (MG) while NASADATA runs perform better based on normalized mean square error (NMSE) and fractional bias (FB). The lack of consistency between these statistical metrics implies there is no clear better source of surface data parameters for this dispersion evaluation study and that both sets of runs exhibit the same general level of dispersion prediction skill.  Predicted peak air concentrations are highly sensitive to surface roughness and displacement height values with CONTROL run peak concentrations between 35 and 49% lower than NASADATA peak simulated concentrations.  Centerline air concentrations in rural areas downwind of an urban release location are marginally sensitive to the choice of surface roughness data with most CONTROL and NASADATA prediction differences within 10%. However, the JU2003 data do not extend far enough downwind for differences in rural surface roughness values to be fully evident. One dispersion test case examined demonstrates a particular sensitivity to roughness data and warrants additional analysis.  The CONTROL runs in this evaluation study use surface characteristics derived from OKC building data that have been shown to give excellent results for JU2003 dispersion studies (Delle Monache 2009). Therefore, the NASADATA runs are being compared against a model configuration already shown to perform exceedingly well for the JU2003 test cases. The 3 comparable dispersion prediction skill shown by the CONTROL and NASADATA runs support the conclusion that GSFC and SSC have developed a robust methodology for determining surface characteristics from NASA satellite data. Recommendations Based on the data evaluation study, we conclude that GSFC and SSC have developed a robust methodology for determining surface roughness. The new satellite-derived surface data capability has the potential to improve numerical weather prediction (NWP) accuracy and dispersion modeling, especially in modeling peak concentrations. Additional evaluation of the impacts of GSFC/SSC surface data is needed to fully assess the locations and scenarios where such surface data are likely to provide the greatest value. Specifically, the additional follow-on studies are recommended:  Evaluation of GSFC/SSC surface data impact in urban centers with different building geometries and densities to confirm the results from this project  Investigation of impacts due to seasonal surface roughness variation particularly for rural areas  Study the impact of GSFC/SSC data on dispersion model concentrations over larger geographic scales where the uncertainty associated with surface data values is expected to have a greater impact than on near-source plume concentrations 4