• Open Access

Improved simulation of the terrestrial hydrological cycle in permafrost regions by the Community Land Model


Corresponding author: S. Swenson, Climate and Global Dynamics Division, National Center for Atmospheric Research, PO Box 3000, Boulder, CO 80307, USA. (swensosc@ucar.edu)


[1] Plausible predictions of future climate require realistic representations of past and current climate. Simulations of the distribution of permafrost in the 21st century made with the Community Climate System Model (CCSM4) indicate that substantial decreases in permafrost extent can be expected, especially under high emissions scenarios. One of the implications of permafrost loss is the potential release of carbon from newly thawed soils into the atmosphere, thus raising its concentration of greenhouse gases and amplifying the initial warming trend. However, the biogeochemical cycle simulated by CCSM4 presents significant biases in carbon fluxes such as gross primary production, net primary production, and vegetation carbon storage in permafrost regions. The biases in the carbon cycle simulated \by CCSM4 are in part due to excessively dry soils in permafrost regions. In this study, we show that the CCSM4 dry soil bias results from the model's formulation of soil hydraulic permeability when soil ice is present. The calculation of the hydraulic properties of frozen soils is first modified by replacing their dependence on total water content with liquid water content only. Then an ice impedance function having a power-law form is incorporated. When the parameterization of the hydraulic properties of frozen soil is corrected, the model simulates significantly higher moisture contents in near-surface soils in permafrost regions, especially during spring. This result is validated qualitatively by comparing soil moisture profiles to descriptions based on field studies, and quantitatively by comparing simulated hydrographs of two large Siberian rivers to observed hydrographs. After the dry soil bias is reduced, the vegetation productivity simulated by the model is improved, which is manifested in leaf area indices that at some locations are twice as large as in the original model.