Daily soil temperature and thaw depth for the entire Arctic terrestrial drainage area are simulated using a one-dimensional heat transfer model with phase change. Analyses of temperature trends at the soil surface and at 2 m depth are presented for the 23-year time period 1980 through 2002. Soil warming is simulated for all permafrost regions, but is most pronounced (0.044°C/yr) at the surface in the continuous permafrost region. Trends for most recent years (1994–2002) are about three times higher. Active layer depth increases significantly for parts of Alaska and northern Canada, and southern and eastern Siberia. As assessed for the major river drainages, the most dramatic active layer deepening occurs in the Yenisey basin.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The past several decades have seen pronounced changes in the Arctic region [Serreze et al., 2000]. While the most easily observable trends include generally rising surface air temperatures and retreat of the sea ice cover, numerous studies have documented changes in the soil thermal regime and permafrost extent. Available observations suggest that permafrost surface temperatures in northern Alaska have increased by 2–4°C in the 20th century [Lachenbruch and Marshall, 1986] and that permafrost extent has declined [Anisimov et al., 2001]. Areas that previously saw cooling of soils now show warming (e.g., up to 2°C in the last decade in eastern Canada) and trends generally seem to be accelerating [Nelson, 2003].
 Soil temperature and active layer depth (ALD, the maximum depth of seasonal thaw) in permafrost regions are useful climate indicators in that they integrate processes that occur at and above the ground surface, such as surface air temperature, snow cover, precipitation, and vegetation. Recent studies implicate deeper active layers and/or thawing permafrost as contributing to increases in winter discharge from Arctic-flowing rivers [Serreze et al., 2002; Ye et al., 2003]. Permafrost warming and active-layer deepening may alter the distribution and growth rate of vegetation, alter sources and sinks of soil organic carbon currently sequested in the upper permafrost, and, in turn, contribute to a positive feedback in the context of global warming. Changes in the soil thermal regime and thaw settlement can create widespread hazards by disrupting roads and other built structures, and through slope instability.
 A major obstacle in assessing spatio-temporal patterns of change is that point and areal measurements of soil temperature and ALD are sparse [Zhang et al., 2001]. Modeling represents an attractive and complementary approach. Here, we present results from simulations of soil temperature and ALD over the 23-year time period 1980–2002. Results are examined over the pan-Arctic terrestrial drainage (the region draining into the Arctic Ocean (Figure 1)) for areas hereon defined as (1) continuous permafrost, (2) discontinuous permafrost, (3) isolated permafrost, and (4) seasonally frozen ground outside the permafrost regions. Respectively, these comprise 42.5%, 10.6%, 17.6% and 20.8% of the pan-Arctic drainage (the remaining 8.5% not addressed here comprises ice sheets, glaciers and lakes). Summaries are also included for major watersheds.
2. Thermal Soil Model
 We use a finite difference model for one-dimensional heat conduction with phase change [Goodrich, 1982] that has been shown to provide excellent results for soil temperatures and active layer depth when driven with well-known boundary conditions and forcing parameters at specific locations [Zhang et al., 1996]. A detailed description of the model is given by Goodrich  and Zhang et al. . A complete description of the model setup for circum-Arctic studies is given by Oelke et al.  and Oelke and Zhang . The model is run one-dimensionally and assumes no lateral heat transfer between the nearly 40,000 25 km × 25 km grid cells defining the pan-Arctic domain. Soil is divided into three major layers (0–30 cm, 30–80 cm, and 80–3000 cm) with distinct thermal properties for frozen and thawed soil, respectively. Calculations are performed on 63 model layers ranging from a thickness of 10 cm for the top 80 cm of soil, to 2 m at 30 m depth, the lower model boundary. The model is spun up for 52 years in order to obtain steady-state initial conditions for temperatures at all model layers. Simulations are performed with a daily time step for the period 1980 through 2002.
 The main forcing parameters are surface air temperature and snow depth at 25 km × 25 km resolution. As daily surface-based measurements in the remote Arctic regions are sparse, we use reanalyzed and remotely-sensed input. Surface air temperatures are from the NCEP/NCAR reanalysis [Kistler et al., 2001]. Following Oelke et al. , a topography correction of surface air temperature is performed using the difference in NCEP and Digital Elevation Model topography on the 25 km × 25 km grid array, in conjunction with temperature lapse rates derived from NCEP tropospheric data. As elevation is a first-order determinant of the spatial variation in surface air temperature, these adjustments effectively improve the resolution of the NCEP data, making them more compatible with the snow-cover data sets. These surface air temperatures are used over snow whereas a correlation expression with soil surface temperature [Zhang et al., 1996] is applied for snow-free surfaces taking into account parameters influencing the surface energy balance, and low vegetation.
 Snow water equivalent (SWE) is derived using re-gridded weekly-averaged passive microwave data from the SMMR (1980–1987) [Chang et al., 1987] and SSM/I radiometers (1987–2002) [Armstrong and Brodzik, 2001]. Snow height is derived from SWE values by dividing by a climatological snow density at the given location and time of year. These snow densities are obtained from a 45-year time series of Canadian snow data [Meteorological Service of Canada, 2000] to define the climatological annual cycles of snow density for tundra, taiga, prairie, alpine and maritime regions [Oelke et al., 2003].
 Very thin snow cover often cannot be detected by passive microwave remote sensing. Therefore, we also use the NOAA-NESDIS weekly snow charts [Armstrong and Brodzik, 2002] that are based on information from several visible-band satellites. For grid cells where the radiometer does not detect snow but the NOAA charts do, we assume a snow thickness of 3 cm. The NOAA charts are most useful at the beginning of the winter season and for the southern margin of snow cover.
 Further model input parameters include soil bulk density, the relative compositions of clay/silt and sand/gravel, and soil water content. Bulk density and fine-grained and coarse-grained soil concentrations for each of the three major model layers are obtained from the IGBP SoilData System [Global Soil Data Task, 2000]. These are used to calculate soil thermal conductivity for frozen and thawed states according to Kersten , modified by the thermal conductivity of peat [De Vries, 1963] for up to 80 cm of soil [Oelke and Zhang, 2004]. Daily soil water content is based on a 20-year model climatology from the University of New Hampshire Permafrost/Water Balance Model. Vegetation cover is only included as part of the surface temperature formulation for tundra areas, but long-term changes of cooling (summer) and insulating effects (winter) may modify soil temperature and ALD trends.
3. Soil Temperature Trend Analysis
 In basic accord with available station records, NCEP surface air temperature shows the strongest annual trend for the 23-year period 1980–2002 over continuous permafrost regions with 0.036°C/yr, and 0.093°C/yr for the period 1994–2002. Snow height has a minimum in the early 1990s, with higher values before and after. Nevertheless, for 1980–2002 a negative snow depth trend is evident for all regions.
 The time series of annual average soil temperatures for the surface (Figure 2a) and 2 m (Figure 2b, updated from Oelke and Zhang ) also reveal the strongest 23-year trends for regions of continuous permafrost (0.044°C/yr and 0.033°C/yr, respectively). Most of the warming occurs after 1994, resulting in much higher trends for this period of 0.142°C/yr (surface) and 0.113°C/yr (2 m depth). For continuous permafrost (Figure 3) most of the soil warming over 1994–2002 occurs during winter (surface: 0.237°C/yr, 2 m: 0.164°C/yr) with smaller trends for spring (0.124°C/yr, 0.151°C/yr), summer (0.089°C/yr, 0.060°C/yr), and autumn (0.119°C/yr, 0.075°C/yr). The spring trend at 2 m depth is almost as high as the trend for winter.
 Discontinuous and isolated permafrost regions (2 and 3) reveal weaker annual surface soil warming trends between 1980 and 2002 (0.032°C/yr and 0.015°C/yr) (Figure 2a). This may be due to large latent heat requirements for soil with temperatures close to 0°C. Seasonally frozen ground regions (region 4) show a weak (and non-significant) cooling trend of −0.002°C/yr. The values for 2 m depth (Figure 2b) are 0.021°C/yr, 0.013°C/yr, and −0.006°C/yr. Trends for 1994 through 2002 are also positive and slightly higher than for the time series as a whole.
4. Trends in Active Layer Depth
 Here we examine the spatial pattern of trends along with summaries for major individual watersheds. The annual ALD is determined from daily values of thaw depth between 1 January and 31 December. ALD for continuous and discontinuous permafrost regions in the Yenisey basin (Figure 4a) has a 23-year average of 142 cm with a strong, statistically significant trend (at the 99% confidence level) of 0.81 cm/yr. The Mackenzie basin (Figure 4b), which has about the same average ALD, also shows a highly significant trend of 0.47 cm/yr. Those trends for the Lena basin (not shown) are 0.35 cm/yr (95% confidence level) whereas the small permafrost area of the Ob basin reveals no significant trends.
Figure 5 a gives the ALD distribution for continuous and discontinuous permafrost regions in 1980. Figure 5b shows the 23-year linear trends that are significant at the 95% confidence level. Areas in southern Alaska, near the Mackenzie River, the Canadian Arctic Archipelago, and south-central Siberia have simulated trends of a few cm/yr. Values in remote areas cannot be confirmed because of lacking observations.
 The ALD is usually reached in late summer or autumn, but can be as early as day 175 (late June) for the northern Canadian Arctic Archipelago, and northern and eastern Greenland, or as late as day 300 (late October) in western Alaska and southern central Canada (Figure 6a). Not surprisingly, the spatial patterns of 23-year linear trends (Figure 6b) is characterized by an increase in the length of thawing season in generally the same areas where the ALD increases significantly. Here, the ALD is reached continually later in the year.
 A number or recent studies have remarked upon recent trends toward increased winter discharge from the major Siberian rivers [see Ye et al., 2003, and references therein]. Part of this change appears to relate to the direct human effects of diversions and impoundments, but increased winter precipitation and surface air temperature appear to have played a role. Part of the temperature link may involve increased snowmelt in the southern headwaters of these river systems. Another explanation involves changes in ALD. Recent winter runoff trends examined by Serreze et al.  are largest in the Yenisey, where observed trends in both ground temperatures and in simulated ALD are most pronounced. With a greater ALD, autumn freeze-up of the active layer would occur later, contributing to more groundwater movement into river channels in autumn, seen as increased winter discharge. The possibility that greater ALD could alter summer streamflow by allowing for greater infiltration, delaying the runoff response to precipitation events, remains to be explored.
 This study was supported by NSF/ARCSS grants OPP-9907541, OPP-9910315, OPP-0229651, and OPP-0229766, and NASA grant NAG5-6820.