A13C-0337
Intercomparison of atmospheric reanalysis data in the Arctic region: To derive site-specific forcing data for terrestrial models

Monday, 14 December 2015
Poster Hall (Moscone South)
Junko Mori, NIPR National Institute of Polar Research, Tokyo, Japan
Abstract:
An intercomparison project for the Arctic terrestrial (physical and ecosystem) models, GTMIP, is conducted, targeting at improvements in the existing terrestrial schemes, as an activity of the Terrestrial Ecosystem research group in the Arctic of Japan GRENE Arctic Climate Change Research Project (GRENE-TEA). For site simulations for four GRENE-TEA sites (i.e., Fairbanks/AK, Kevo/Finland, Tiksi and Yakutsk/Siberia), we needed to prepare continuous, site-fit forcing data ready to drive the models. Due to scarcity of site observations in the region, however, it was difficult to make such data directly from the observations. Therefore, we decided to create a backbone dataset (Level 0 or Lv0) first by utilizing the reanalysis data to derive the site-specific data (Level 1 or Lv1).

For selection of the best dataset for our purpose, we compared four atmospheric reanalysis datasets, i.e., ERA Interim, JRA-55, NCEP/NCAR Reanalysis 1, and NCEP-DOE Reanalysis 2, in terms of the climatic reproducibility (w.r.t. temperature at 2 m and precipitation) in the region north of 60°N. CRU for temperature and GPCP for precipitation were also used for monthly-mean ground-level climate. As we will show ERA-Interim showed the smallest bias for both the parameters in terms of RMSE. Especially, air temperature in the cold period was reproduced better in ERA-Interim than is in JRA-55 or other reanalysis products. Therefore, we created Lv0 from ERA-Interim. Comparison between the site observations and Lv0 showed good agreement except for wind speed at all sites and air temperature at Tiksi, a coastal site in the eastern Siberia. Air temperature of ERA-Interim showed significantly continental characteristics while the site observation more coastal. The 34-year-long, hourly, site-fit continuous data (Lv1) for each of the GRENE-TEA sites was then created from the Lv0 values at the grid point closest to the site, by merging with the observations.