Comment on “A projection of severe near‐surface permafrost degradation during the 21st century” by David M. Lawrence and Andrew G. Slater

1 The sensitivity of permafrost to climate warming has been known for a considerable time e.g., Goodwin et al., 1984; Lachenbruch et al., 1988, and there is mounting evidence that recent climate warming is associated with permafrost thaw and consequent terrain effects Payette et al., 2004; Jorgenson et al., 2006. Lawrence and Slater 2005 address the evolution of permafrost in the Northern Hemisphere under climate change predicted during the 21st century using the Community Climate System Model, version 3 CCSM3. Unlike previous models of climate-induced changes in the global extent of permafrost, soil thermal evolution is embedded in CCSM3. The model is primarily an atmosphere-ocean model and restricts consideration of the subsurface thermal regime to the uppermost 3.43 m of the ground. Lawrence and Slater 2005 predict a 60 to 90% reduction in the geographic extent of “near-surface permafrost” by 2100. No previous research has predicted or documented such radical changes in permafrost conditions over so short a period of time. This commentary discusses Lawrence and Slater's model calibration and conclusions from the perspective of permafrost science. 2 Permafrost is defined as “earth materials that remain continuously at or below 0°C for at least two consecutive years” van Everdingen, 1998. Permafrost is therefore a condition of ground climate. The geographic distribution of permafrost is usually represented on continental-scale maps by a series of quasi-latitudinal “zones” representing the dominance of laterally continuous, discontinuous, or sporadic permafrost within them. The basis for this classification is declining ground temperature along a poleward trajectory. Ground temperatures within a geographic region vary about a mean value due to differences in surface conditions. In continuous permafrost, the variation is insufficient to raise ground temperatures above 0°C. In discontinuous permafrost, the spatial mean approaches 0°C, and, accordingly, the proportion of ground underlain by permafrost is smaller. In the sporadic zone, only favorable conditions lead to the formation or maintenance of permafrost. At its southernmost limit in the Northern Hemisphere, the annual mean temperature of permafrost is a fraction of a degree below 0°C, as in southern Alaska and Yukon Territory Burn, 1998. 3 Lawrence and Slater 2005, paragraph 8 state that CCSM3 extent of permafrost “compares favorably” with observed estimates of continuous permafrost. It is inappropriate to compare the extent of all permafrost with the spatial distribution of continuous permafrost alone, because the latter is a more constrained phenomenon than the permafrost produced by CCSM3. Comparison of relatively warm modeled conditions with colder field conditions, the basis of the validation procedure, does not confirm “a reasonable simulation” Lawrence and Slater, 2005, paragraph 17; rather, it introduces an error of up to 40% in the calibration. The extent of contemporary permafrost in CCSM3 is 10.5 million km2, while the observed estimate of discontinuous and continuous permafrost (which corresponds conceptually to the modeled “permafrost”) is between 14.6 and 11.8 million km2 Zhang et al., 2000. The large size of the grid cells employed in the modeling exercise is not justification for ignoring discontinuous permafrost, which underlies vast areas where the mean ground temperature is still below 0°C. 4 Lawrence and Slater 2005, Figure 1d indicate that the global extent of permafrost decreased by about 25% between 1900 and 2000. We know of no data indicating that the extent of continuous permafrost decreased during the 20th century, even though there has been considerable effort to monitor changes in ground conditions e.g. Brown et al., 2000; Romanovsky et al., 2002. There are a few reports that permafrost is disappearing near its southernmost margins e.g., Kwong and Gan, 1994; Jorgenson et al., 2001, but these investigations were conducted hundreds of km from the zone of continuous permafrost, and well outside the area of permafrost mapped by Lawrence and Slater 2005. We know of no field observations indicating that the global extent of permafrost declined substantially (i.e. by 25%) in the last 100 years. In fact, thin permafrost that formed in the peatlands of central Canada during the Little Ice Age has persisted through the last century due to the thermal properties of peat and the presence of “excess ice,” that is, ice whose volume exceeds the porosity of unfrozen soil Halsey et al., 1995. 5 In the Northern Hemisphere, permafrost thickness ranges from less than 1 m at its southernmost margins to more than 1 km in unglaciated parts of Siberia. A relatively thin seasonally thawed active layer lies between the ground surface and permafrost. In many permafrost environments, sediments (and some bedrock) immediately below the base of the active layer are ubiquitously ice-rich. In the field, the ice content in sediments at the base of the active layer varies from saturation to almost 100% by volume. The presence of excess ice is a critical control on the potential increase in active-layer thickness, owing to latent heat effects Shur et al., 2005. In CCSM3, the moisture content in permafrost varies between saturated and unsaturated conditions and is restricted to soil only Lawrence and Slater, 2005, paragraph 6. Because CCSM3's soil characterization neglects excess ice, active-layer thickening is unduly accelerated, and the disappearance of permafrost from the upper 3.43 m is greatly overestimated. 6 The presence of excess ice retards the rate of permafrost degradation. For example, with 50% excess ice, the upper 10 cm of permafrost must be thawed for the active layer to deepen by 5 cm Mackay, 1970. In CCSM3, for every unit increase in the thickness of the active layer, there is unit degradation of permafrost. The absence of thaw consolidation from the model further accelerates the projected disappearance of permafrost from the upper 3.43 m of the ground, because the soil in CCSM3 is of constant thickness. 7 CCSM3 estimates that mean ground temperatures in the Alaskan Arctic are currently between −1°C and 0°C Lawrence and Slater, 2005, Figure 2. As noted in paragraph 2, such mean ground temperature represents conditions associated with discontinuous permafrost near its warmest limit, not the continuous permafrost of Arctic Alaska, where the ground is > 5°C cooler. Permafrost temperatures in this region vary from −3 to −10°C, with the regional average between −5 and −7°C Osterkamp, 2003. Therefore Lawrence and Slater 2005 grossly overestimate the current temperature in permafrost. 8 Lawrence and Slater 2005 presented results indicating that 60 to 90% of “near-surface” permafrost will disappear by 2100. Lawrence and Slater's 2005, paragraph 17 statement that “permafrost is reasonably simulated in CCSM3 20th century climate simulations” is incorrect, because they compare dissimilar objects: the modeled total extent of permafrost and the extent of empirical continuous permafrost. While CCSM3 shows permafrost declining in extent by 25% in the 20th century, no change of such magnitude has been observed. While the extent of CCSM3 permafrost corresponds to observed continuous permafrost, the temperature of CCSM3 permafrost does not. The extent of CCSM3 permafrost poorly reproduces the observed extent of continuous and discontinuous permafrost. The warmer temperatures of present CCSM3 permafrost increase its susceptibility to greater thaw. The lack of consideration of near-surface excess ground ice by CCSM3 similarly inflates the sensitivity of modeled permafrost to thawing. As a result, we consider the results to be gross exaggerations of the physical response of permafrost to climate warming. We note that deep permafrost thaw did not occur during the warmest climatic conditions of the Holocene in northwest Canada. Instead, during a warm interval lasting many hundred years, at sites that are now in warm discontinuous permafrost, the active layer merely doubled over its present thickness e.g., Burn et al., 1986. 9 We thank O. Anisimov, J. Brown, K. M. Hinkel, N. Shiklomanov, S. Smith, and V E. Romanovsky for comments and helpful discussion.