Long-term evaluation of the Hydro-Thermodynamic Soil-Vegetation Scheme's frozen ground/permafrost component using observations at Barrow, Alaska

• Published in: Journal of Geophysical Research. D. Atmospheres 2006-02, Vol.111 (D4), p.n/a
• Main Authors: Mölders, Nicole, Romanovsky, Vladimir E.
• Format: Article
• Language: English
• Subjects: climate modeling; Earth sciences; Earth, ocean, space; evaluation; Exact sciences and technology; permafrost
• ISSN: 0148-0227; 2156-2202
• DOI: 10.1029/2005JD005957

The multi‐layer frozen ground/permafrost component of the hydro‐thermodynamic soil‐vegetation scheme (HTSVS) was evaluated by means of permafrost observations at Barrow, Alaska. HTSVS was driven by pressure, wind, air temperature, specific humidity, snow‐depth, rain, downward shortwave and long‐wave radiation observations for 14 consecutive years. Observed soil temperature data are available at various times during this period. HTSVS predicts soil temperatures that are slightly too low with root mean square errors (RMSEs) of, on average, less than 3.2 K. Sensitivity studies suggest that the treatment of snow and vegetation cover may be reasons for the inaccuracy. HTSVS' original thermal conductivity parameterization provides thermal conductivity values that are too high compared to typical observations. Introducing a parameterization frequently used in the permafrost research community, which was modified for application in numerical weather prediction (NWP) and climate models and model consistency in HTSVS, improves soil temperature predictions and reduces RMSEs in some layers by up to 1 K, and on average by 0.2 K. Assuming five to ten layers for the first 2 or 3 m as is usually done in NWP and climate modeling is insufficient to capture the active layer depth, because the number and position of the grid nodes play a role. The depth of the lower boundary of the soil model and the boundary condition affect the overall performance. Consequently, under current computational possibilities, simulating permafrost and the active layer in atmospheric models requires a compromise between the degree of accuracy and affordable computational time.