Data Assimilation for Arctic System Reanalysis

The Arctic System Reanalysis (ASR), which is approaching the operational stage, includes atmospheric, sea ice and land surface representations. Through atmospheric, land and sea ice data assimilation a broad-based set of historical data streams from the surface and space will be combined with measurements of the physical components. The domain of the ASR includes all of the northward flowing rivers that empty into the Arctic Ocean. The ASR is based upon: the polar-optimized version of the Weather Research and Forecasting model (Polar WRF) , WRF variational data assimilation (WRF-Var) and High Resolution Land Data Assimilation System (HRLDAS). Polar WRF includes an improved Noah land surface model and specifications for the following sea ice attributes: extent, concentration, thickness, albedo and snow cover. WRF-Var assimilates NCEP-PREPBUFR observation data (in-situ surface and upper air data, remotely sensed retrievals and satellite radiance data). HRLDAS is a vital component of ASR that assimilates snow cover and depth, observed vegetation fraction and albedo. The current HRLDAS uses NASA, NESDIS, and NOAA satellite observations to describe these surface properties.

ASR data assimilations with reduced resolution over a single domain (30 km) have been performed from 2000 to 2009 on the Ohio Supercomputer Center's Glenn Cluster; this is known as ASR-Interim. The T255 (0.7 degrees) horizontal resolution ERA-Interim reanalysis surface and upper air model level data are used to provide the background initial and lateral boundary conditions for the ASR Interim. Polar WRF version 3.2.1, WRF-Var version 3.2.1 and HRLDAS are used for the data assimilations.

The annual total accumulated precipitation from the ASR data assimilation and from ERA-Interim are very similar. The added benefit of higher resolution with the ASR-Interim data assimilation (30 km for ASR versus ~80 km for ERA-Interim) can be seen along the mountainous west coast of North America, southeast Greenland, southeast Iceland, and western Scandinavia where more detailed and realistic features are evident in ASR. The results of the ASR data assimilation for 10 years are compared with 3-h surface observations from across the ASR domain. Comparisons of results from NCEP AVN and ERA-Interim with 3-h surface observations are also made. The high skill in resolving surface pressure is seen, along with good skill for 2-m temperature and dew point temperature. The 10-m wind speed skill is much higher than AVN and higher than ERA-Interim during summer. The 2m temperature correlation of ASR is much high than AVN during winter time. ASR has much better performance than AVN and comparable performance with ERA-Interim. These comparisons are based on observations from more than 10,000 surface stations obtained from the National Climatic Data Center (NCDC).

Based upon the very encouraging results with the single domain data assimilation for 10 years (ASR Interim), the ASR team will perform 11 years (2000-2010) with a single domain at 10 km resolution. New versions of Polar WRF, WRF-3DVar improved by NCAR and Noah Land Data Assimilation will be used for the final run.