High Resolution Reanalysis over Complex Terrain with the Climate Four-Dimensional Assimilation System

The Research Applications Laboratory (RAL) and the Atmospheric Chemistry Observations and Modeling (ACOM) Laboratory at the NSF National Center for Atmospheric Research (NCAR) have been funded by the Colorado Department of Public Health and Environment (CDPHE) to conduct a 5-year meteorological reanalysis at high-resolution (Δx = 1 km) over Colorado, a region characterized by complex terrain. This effort supports CDPHE's air quality permit modeling process which involves using plume dispersion models such as AERMOD to assess air pollution concentrations from points sources. The developed dataset will allow CDPHE to better characterize the atmospheric conditions over complex terrain in urban areas than current approaches.

For this application, the Meteorological Assimilation Data Ingest System (MADIS) observations are assimilated into the regional climate version of the WRF-based Real-time Four-Dimensional Assimilation (RTFDDA) known as Climate FDDA (CFDDA). At a given hour, there are more than 4,600 surface observations in the 1-km domain covering the state of Colorado. It is expected that the reanalysis can better represent the local weather and microclimate such as urban heat island effect and mountain-valley circulation. CFDDA with a global version of MM5 has been used previously to create a 0.4o global 20 year reanalysis. Also a WRF-based CFDDA was applied to generate a 5-year 1 km reanalysis in southeastern China.

A two-domain setup is used where the 1-km domain is nested within a 3-km parent domain. The larger 3-km parent resolves larger scale features and provides boundary conditions to the 1-km domain at every time step. Before the production of the 5 year reanalysis, four month-long reanalysis from each of the four seasons will be run as a test with part of the surface observations withheld from assimilation. The withheld observations serve as independent observations for validation b. The methodology to withhold observations for assimilation is as follows. The surface observations are mapped into the parent grid (Δx = 3 km). Provided that there are at least 4 observations within the parent grid cell, one-third of the surface observations (rounded off to the nearest integer) are held for validation. Bias error, mean absolute error, root mean square error of 2-m temperature, 2-m dewpoint, 10-m wind speed and direction will be presented.