Simulation of Present-Day and Future Permafrost and Seasonally Frozen Ground Conditions in CCSM4

The representation of permafrost and seasonally frozen ground and their projected twenty-first century trends is assessed in the Community Climate System Model, version 4 (CCSM4) and the Community Land Model version 4 (CLM4). The combined impact of advances in CLM and a better Arctic climate simulation, especially for air temperature, improve the permafrost simulation in CCSM4 compared to CCSM3. Present-day continuous plus discontinuous permafrost extent is comparable to that observed [12.5 x 10(6) versus (11.8-14.6) x 10(6) km(2)], but active-layer thickness (ALT) is generally too thick and deep ground (>15 m) temperatures are too warm in CCSM4. Present-day seasonally frozen ground area is well simulated (47.5 x 10(6) versus 48.1 x 10(6) km(2)). ALT and deep ground temperatures are much better simulated in offline CLM4 (i.e., forced with observed climate), which indicates that the remaining climate biases, particularly excessive high-latitude snowfall biases, degrade the CCSM4 permafrost simulation. Near-surface permafrost (NSP) and seasonally frozen ground (SFG) area are projected to decline substantially during the twenty-first century [representative concentration projections (RCPs); RCP8.5: NSP by 9.0 x 10(6) km(2), 72%, SFG by 7.1 x 10(6), 15%; RCP2.6: NSP by 4.1 x 10(6), 33%, SFG by 2.1 x 10(6), 4%]. The permafrost degradation rate is slower (2000-50) than in CCSM3 by similar to 35% because of the improved soil physics. Under the low RCP2.6 emissions pathway, permafrost state stabilizes by 2100, suggesting that permafrost related feedbacks could be minimized if greenhouse emissions could be reduced. The trajectory of permafrost degradation is affected by CCSM4 climate biases. In simulations with this climate bias ameliorated, permafrost degradation in RCP8.5 is lower by similar to 29%. Further reductions of Arctic climate biases will increase the reliability of permafrost projections and feedback studies in earth system models.