Maximum annual development of the active layer above permafrost has been monitored at seven 1-km2 Circumpolar Active Layer Monitoring (CALM) sites in northern Alaska since 1995. Grid nodes are spaced at 100-m intervals, yielding a regular array of 121 (11 × 11) data collection points. Three sites are located in the Arctic Foothills physiographic province, and four are on the Arctic Coastal Plain. Air and soil temperature measurements are made at each site, and soil moisture is monitored at most. Six years of record permit several general conclusions: (1) At the landscape scale, end-of-season thaw depth is strongly correlated with local air temperature on an interannual basis. All sites experienced maximum average thaw depth in 1998 and a minimum in 2000, consistent with the warmest and coolest summers during the period of record. The active layer, however, may exhibit Markovian behavior over multidecadal periods. (2) There is significant intrasite variation in thaw depth and near-surface soil moisture content within each 1-km2 grid, reflecting the local influence of vegetation, substrate properties, snow cover dynamics, and terrain. (3) On the Coastal Plain, thaw depth is significantly greater in drained thaw-lake basins, resulting in a bimodal distribution of thaw depth related to primary landscape elements. (4) Foothills sites demonstrate large spatial and interannual variability resulting from microtopography and temporal variations of soil moisture content, making predictive mapping of thaw depth problematic at the scale and resolution of the grids. The spatial pattern of thaw depth across sites on the Coastal Plain is relatively consistent, although lake margins exhibit more complex patterns attributable to fluctuating water levels.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The thickness of the active layer, or near-surface layer of Earth material above permafrost undergoing seasonal freezing and thawing, is of considerable importance in high-latitude environments because most hydrological, biological, and biogeochemical activity takes place within it [Kane et al., 1991]. Under a warming climate, thickening of the active layer could release water and carbon currently sequestered in the upper permafrost, contributing to positive feedback effects in the context of global warming. Widespread subsidence of the ground surface may occur in areas in which near-surface permafrost is ice-rich, affecting drainage patterns, ecological relationships, and human infrastructure [Beilman et al., 2001; Jorgenson et al., 2001; Nelson et al., 2001].
 Detailed, temporally integrated regional estimates of geocryological parameters present a major challenge to global-change research, owing to: (1) the relative inaccessibility of much of the Arctic; (2) the great local variability of thaw depth; and (3) the current necessity for field-based observation programs. Until effective remote-sensing methods are developed to measure maximum seasonal thaw depth, the dual issues of spatial and temporal variability of the active layer can best be assessed through repeated observation at representative plots over long time periods. The “scale-up” approach to spatial integration [Root and Schneider, 1995] can be highly effective for active-layer mapping given adequate data on climate and surface and subsurface properties [Nelson et al., 1997; Shiklomanov and Nelson, 1999; Klene et al., 2001].
 The Circumpolar Active Layer Monitoring (CALM) program developed over the last decade as a leading edge in comprehensive efforts to study the impacts of climate change in permafrost environments. Begun as an effort to recover data from defunct observation programs, to reactivate some of those programs, and to involve a limited number of new locations, CALM grew rapidly under the umbrella of a related biological climate-change program [Nelson et al., 1996; Arft et al., 1999]. Eventually, CALM attracted a quasi-independent source of support and established close linkages with international monitoring programs [Burgess et al., 2000]. At this writing, the program consists of about 85 active sites in 11 countries in the Northern Hemisphere. A detailed historical treatment of CALM is provided by Brown et al. .
 This paper describes results obtained during the first 6 years (1995–2000) of a major component of the CALM program: spatial time series of active-layer thickness from a network of seven 1-km2 observation sites on the North Slope of Alaska. These grids form the core of CALM in northern Alaska, and many of the program's protocols and sampling conventions were developed on them. Records from other CALM sites with long records are reviewed by Brown et al. .
2. Regional Background
 CALM's initial focus was on two series of sites in northern Alaska, both of which were derived in part from antecedent observation programs. The regional network is comprised of two north-south transects across the North Slope (Figure 1), oriented along the regional climatic gradient [Clebsch and Shanks, 1968; Haugen, 1982]. The Western Transect consists of sites at Barrow and Atqasuk on the Coastal Plain, and Ivotuk in the Arctic Foothills. Because the latter site was established much later than the others, this paper treats only the Barrow-Atqasuk portion of the transect. The Kuparuk transect, 340 km to the east, parallels the Dalton Highway and the Trans-Alaska Pipeline System (TAPS) from Prudhoe Bay on the Coastal Plain to Toolik Lake in the northern foothills of the Brooks Range. ARCSS/CALM grids are located within the Kuparuk River basin at Betty Pingo and West Dock on the Coastal Plain, Happy Valley in the Sagwon Upland, and Toolik Lake and Imnavait Creek in the Arctic Foothills. ARCSS/CALM sites along the Kuparuk Transect grew out of several research programs, including U.S. Army Cold Regions Research and Engineering Laboratory (CRREL) activities in the Prudhoe Bay Oil Field [Brown, 1975; Walker et al., 1980], the U.S. Environmental Protection Agency's R4D program [Reynolds and Tenhunan, 1996], and scientific investigations at the University of Alaska's Toolik Lake Field Station. The five grids comprising the Kuparuk transect were in place by 1994, and were used intensively by the Arctic Flux Study [Kane and Reeburgh, 1998].
3.1. Field Procedures
 The CALM program uses a wide variety of site characteristics, data-collection procedures, sampling designs, and instrumental observations, adapted to meet local circumstances and the analytic goals of various investigators. On Alaska's North Slope, thaw depth is determined through replicated measurement at 121 nodes on seven 1-km2 grids, each consisting of precisely surveyed stakes anchored in ice-rich permafrost at 100-m horizontal intervals [Nelson et al., 1998b]. Thaw depth is determined by inserting a 1-cm diameter graduated steel rod into the soil to the point of refusal. Probing is carried out in mid- to late August, when thaw is near its seasonal maximum.
 At some grid nodes it was not possible to measure thaw depth. This situation arises when a grid node is located in deep ponds or lakes, streams, or a stony substrate. In some cases, grid nodes coincide with cultural features such as roads and gravel pads. These missing data points were not included in the calculation of summary statistics or in the data analysis.
 Large interannual variability at individual grid nodes can result from three influences: (1) when ponding at the grid node occurs in some years but not others; (2) when there is a high degree of lateral thaw variability over short distances (<1 m); and (3) when the substrate is stone-rich, yielding erroneous probing estimates of thaw depth. The latter is particularly problematic in glacial tills.
3.2. Analytic Procedures
 The maps presented in this paper utilize a 21 × 21 matrix generated using a linear interpolation algorithm [Golden Software, 1999]. This procedure yields relatively smooth surfaces with readily interpretable visual patterns. Data used for statistical analyses were not preprocessed.
 Thaw depth at individual grid nodes is forced by air temperature, but is also influenced by such local factors as vegetation and soil properties, some of which vary interannually [Miller et al., 1998]. To examine spatial variability over the 6-year time series, we computed a normalized index of variability (Iv), given by:
where Zavg is the areally averaged thaw depth for a particular year and Zi is the node-specific value. Thus, if in a particular year when the areally averaged thaw depth is 50 cm and the node-specific thaw depth is 55 cm, the normalized local thaw depth Iv exceeds the grid average by 10%. If greater thaw occurs the following year, with an average of 60 cm for example, we would expect the grid node to have a thaw value of 66 cm if responding linearly to thermal forcing alone.
 If other influences are held constant, the node-specific thaw depth should respond relatively consistently from year to year. Over the 6-year period of record, for example, the normalized thaw depth at a grid node may range from a minimum of 12% to a maximum of 15%. In this case, the interannual node variability (INV) is 3%, which indicates a high degree of consistency over the period of record. Conversely, a grid node may demonstrate high thaw variability. In one year, the normalized thaw depth may be 20% greater than the grid mean for that year, whereas in another year it may be 30% less than the annual grid average. Thus, the range over the 6-year period of record is 50%, meaning that the node demonstrates a high degree of interannual variability (INV = 50%) in response to site-specific factors. We can also calculate a grid-averaged INV to determine the response consistency over the period of record, and map the INV to determine the pattern of thaw consistency as it varies across the grid.
4. Site Descriptions and Results
4.1. Coastal Plain Sites
 The first four sites are situated in the Arctic Coastal Plain physiographic province [Wahrhaftig, 1965]. Their common characteristics include low relief, thaw lakes and drained or partially drained thaw lake basins, and moist to wet acidic tundra (Table 1).
Glacial deposits, stream, floodplain, water tracks
moist & wet acidic tundra
Typic Aquiturbel (U), Typic Hemistel (L)
295 to 342
Glacial deposits, stream, floodplain, water tracks
moist acidic tundra
 The Atqasuk site is located along the Meade River on the inner Arctic Coastal Plain, approximately 100 km inland. The 1-km2 grid is situated on an ancient sand sea [Everett, 1980; Carter, 1981] developed on a 1% north-facing slope. A lake occupies a portion of the northeastern quadrant, and the well-drained sandy higher slopes are characterized by moist acidic tundra.
 A 1 × 1 km2 grid of precisely surveyed stakes at 100-m intervals was established at Atqasuk in 1995 under the auspices of the U.S. National Science Foundation's Arctic System Science (ARCSS) program. The ARCSS/CALM grid includes the vicinity at which measurements were made in the RATE program during the 1970s [Batzli, 1980]. Simultaneous measurement of air and shallow (1 m) ground temperature has been operating continuously at the site since 1994. The edaphic environment at Atqasuk is distinct from that of other sites in the Alaska CALM network.
 End-of-season thaw depth patterns for the period 1995–2000 are shown in Figure 2a. These and subsequent maps (Figures 2 and 3) share a common depth scale to facilitate visual comparison, and the mean for each year is shown in parentheses. Areas without contours are blank nodes in which data were not collected owing to the presence of the lake.
 An additional map was generated using the average thaw depth at each grid node over the 6-year period of record. This map of average thaw (Figure 3a) is shown adjacent to a terrain map to demonstrate the correspondence between thaw-depth patterns and primary terrain features. At Atqasuk, relatively shallow thaw occurs on the well-drained sandy uplands. The thaw depth increases in the lake basin, with maximum values recorded in the wet margin of the lake. Saturated soil conditions enhance heat flow to depth by increasing the bulk thermal conductivity of the upper substrate.
 The final map in Figure 3a illustrates the pattern of thaw variability. This map was generated using the node-specific range of annual INVs over the 6-year period. At Atqasuk, thaw depth in the uplands is relatively consistent. By contrast, lowlands surrounding the lake basin experience high interannual variability. Overall, the INV average for the grid is 28%. This variability map allows us to estimate the magnitude of thaw variation and to identify areas or grid nodes where thaw depth is not responding consistently to forcing. At Atqasuk, fluctuations in lake level and soil moisture content probably cause the higher thaw variability around the lake margin.
 A 1 × 1 km ARCSS/CALM grid was installed at Barrow in the early 1990s to facilitate collection of ecological, edaphic, and geocryological information at standardized locations that could be tied into coordinate systems with high-order accuracy. Simultaneous air and shallow (1 m) ground temperatures observations have been made at several locations on the Barrow grid since 1993. The environment at Barrow was described in a series of papers describing the work of the Tundra Biome Program during the 1970s [Brown et al., 1980].
 The Barrow 1-km2 grid is located about 5 km east of the village of Barrow on the outer Arctic Coastal Plain. The site is situated on reworked silts of marine origin (Table 1). As shown in Figure 3b, terrain occupied by the grid encompasses a drained lake basin or lagoon to the west (Central Marsh). To the east is a polygonized “upland,” a term used here to convey relative elevation within the low-relief Coastal Plain. These areas are separated by a north-south trending beach ridge composed of sandy gravels. The polygonized upland is occupied by moist acidic tundra, with well-developed low-centered ice-wedge polygons in the northeast and high-centered ice wedge polygons to the southeast. The latter developed in response to thermomechanical incision by streams draining to Elson Lagoon in the southeastern corner of the grid; ephemeral flow is confined to the ice-wedge troughs.
 Annual end-of-season-thaw depth patterns are shown in Figure 2b, and depict a high degree of spatial uniformity across most of the grid. The generally shallow average thaw depth reflects the colder summer climate at this coastal location as compared to Atqasuk. Enhanced thaw is associated with the beach ridge.
 The 6-year average thaw depth in Figure 3b demonstrates the relation between thaw depth and terrain. The active layer is generally thicker on the west-facing beach ridge slope, with some areas exceeding the grid mean by a factor of two. The grid average INV (27%) is similar to that at Atqasuk; maximum variations are associated with those nodes of maximum thaw development.
 A Vitel Hydra® soil moisture probe was employed to measure volumetric soil moisture content (%) at the grid nodes for the near-surface layer (0–7 cm). The pattern for two years is shown in Figure 4, and clearly demonstrates the contrast between Central Marsh and the beach ridge, the areas of low- and high-centered polygons, and the difference between a “normal” precipitation year (1996: 46 mm in summer) and an extremely wet year (1997: 127 mm).
 Snow depth was measured each spring on the Barrow grid. A representative sample from early May 1998 is also shown in Figure 4. Prevailing winds in winter are from the east, encouraging development of a drift to the lee of the beach ridge. Areas of maximum snow accumulation coincide exactly with those of maximum thaw depth. Drifts often persist several weeks after the snow cover has ablated from the surrounding tundra, and thus serve to delay the onset of soil thaw. The insulating effect of deep snow inhibits heat loss in winter, however, so less heat energy is required to thaw the soil in spring. Temperature loggers installed at the base of the snowdrift indicate minimum winter soil surface temperatures 7–10°C warmer than in terrain without significant drifting. Snow-depth patterns, as affect by topography, therefore have an influence on the spatial patterns and magnitude of thaw. This effect may also be influenced by the permeable beach ridge gravel substrate; rapid infiltration, and lateral downslope flow of soil water at the base of the active layer, enhances thermal ablation.
4.1.3. Betty Pingo and West Dock
 The Betty Pingo and West Dock grids are located on the outer Arctic Coastal Plain within the Prudhoe Bay oil field. Contrasts between polygonized “uplands” and drained thaw-lake basins constitute the primary relief at these sites, and lakes and ponds are ubiquitous elements of the landscape (Figures 2c and 2d). Landcover units include moist nonacidic tundra and wet tundra vegetation, and Typic Aquorthel soils developed in alluvium [Walker and Bockheim, 1995].
 The magnitude and pattern of active-layer thickness on the Betty Pingo grid over the period of record are shown in Figure 2c. Average thaw depth is shown in Figure 3c. Shallow thaw depths are found on the flat uplands separating the basins. There are two areas of enhanced thaw. One is clearly associated with the wet, low-lying basin. The second is apparent in the southeastern quadrant, where a west-facing slope exists between the higher ground and the ponded upland. Although snow depth measurements have not been collected from this site, the pattern is similar to that observed at Barrow. It therefore appears that snow drifting at the break-in-slope has resulted in greater thaw depths. Interannual node variability is maximized in these two areas, but the grid-averaged INV is relatively small (22%).
 West Dock is located about 10 km from the Betty Pingo site and is very near the coast. The landscape, vegetation, and soils are similar at the two sites. Two drained thaw-lake basins occupy the eastern half of the grid and are separated from the upland to the west by a prominent beach ridge. A network of low-centered ice-wedge polygons is developed on the ponded tundra upland. Water depth in the polygon centers depends primarily on the amount of precipitation-both winter snow cover thickness and summer rainfall. Thaw patterns for the period 1995–2000 are shown in Figures 2d and 3d. Greater thaw depths occur in the wet drained lake basins and near large ponds, while the ponded polygonized uplands experience generally shallow thaw. Higher INV values are associated with the surface depressions and ponds, but the grid average is relatively small at 18%.
4.2. Discussion of Coastal Plain Sites
 At Coastal Plain sites, the frequency distribution of thaw-depth values is bimodal, with wet thaw-lake basins and lake margins experiencing significantly greater thaw than the “uplands.” This distribution is not revealed by the grid average or measurements of central tendency, but can be verified by separating measurements collected from the two areas for statistical comparison.
 Because the distribution of the measurements collected in the basins and uplands are distinctly nonnormal, nonparametric techniques were used. The Mann-Whitney test compares sample medians, whereas the Kolmogorov-Smirnov test focuses on maximum pairwise differences between the cumulative frequency distributions of the samples. Summary statistics and test results for the 1995–2000 grid node averages are shown in Table 2. They clearly demonstrate that at Atqasuk, West Dock, and Betty Pingo, thaw depth in the basins is significantly greater then in the uplands. Comparison on an annual basis would yield similar results.
Table 2. Summary Statistics for Upland and Basin Grid Nodes at Coastal Plain Sitesa
Nonparametric Mann-Whitney (M-W) and Kolmogorov-Smirnov (K-S) tests results reported as significant (s) or not significant (ns). p = significance level.
p = 0.204
p = 0.000
p = 0.003
p = 0.000
p = 0.466
p = 0.000
p = 0.007
p = 0.000
 The exception to this pattern is Barrow. Thaw depth in the upland is comparable to the upland at Atqasuk, but the basin (Central Marsh) does not experience significantly greater thaw depths. This may be explained by the fact that, although the soil is wet, there is no permanent standing water near the eastern margin of Central Marsh. By contrast, the other Coastal Plain sites have permanent bodies of water. Examination of the maps of average thaw depth for these sites (Figures 3a, 3c, and 3d) reveals a general increase in thaw depth from the basin margins toward the basin center. At Atqasuk and Betty Pingo in particular, lake margins are often covered by shallow standing water, resulting in average thaw depths in the basins being 40% (Betty Pingo) and 95% (Atqasuk) greater than observed in the uplands, and demonstrating the strong influence of standing water depth on thaw depth.
4.3. Foothills Sites
 These sites are situated near the Dalton Highway in the Arctic Foothills physiographic province [Wahrhaftig, 1965] south of Prudhoe Bay. The area was glaciated during the Pleistocene [Hamilton, 1986], and the surface deposits are largely glacial till with a discontinuous loess cover [Walker et al., 1998]. Terrain is much more variable in the Foothills grids than in those on the Coastal Plain.
4.3.1. Toolik Lake
 This grid occupies the flanks of a bedrock-cored hill adjacent to Toolik Lake. Moist acidic tundra has developed atop glacial tills [Auerbach et al., 1996]. Vegetation and soils vary with slope, aspect, and drainage, and are discontinuous.
 Thaw patterns and magnitude are shown in Figure 2e, and average thaw depth in Figure 3e. Thaw is shallow along steeper slope facets near the base of the hill, increases up the north- and east-facing slopes, and achieves a maximum near the top of the rounded hill.
 The general pattern of active-layer thickness on the Toolik grid appear to be related to details of the local topography. The upper portions of the hill are comprised of gentle, concave slopes with higher soil moisture than the steeper, convex slopes near the base. Moreover, receipts of direct solar radiation are higher around the hill top and decrease downslope on the north-facing slope. Topography therefore exerts a direct influence on the general thaw pattern [cf. Nelson et al., 1997].
 There is, however, a great deal of local interannual variation; the grid INV average of 40% is nearly twice that experienced at the Coastal Plain sites. Variation appears to be uniformly distributed across the grid, although higher magnitude INVs occur along the lake margin and in some of the wetter areas.
4.3.2. Imnavait Creek
 The Imnavait Creek site encompasses a north-south trending stream valley, much of the valley's slopes, and an adjoining hill crest. Wet acidic tundra with a thick organic mat occupies the valley bottom. The higher elevations are occupied by moist acidic tundra, but the slopes are broken by numerous water tracks that become more prominent downslope [McNamara et al., 1999]. These narrow zones contain shrub vegetation, especially willow.
 Patterns of thaw for the period 1995–2000 are shown in Figure 3f. No interannual pattern is readily discernible. Furthermore, average thaw does not appear to be related to topography. There is, however, a distinct tendency for patterns to be oriented downslope, with enhanced thaw associated with water tracks. The average INV is high at this site (37%) but no obvious correlation with topography or landscape features is apparent.
4.3.3. Happy Valley
 The Happy Valley site occupies gentle, loess-covered hills with water tracks and a stream in the southeast corner. The vegetation is moist acidic tundra, with shrubs (willow) in the water tracks and riparian zone.
 The annual patterns of thaw, as demonstrated in Figure 2g, are quite varied. The influence of the terrain on thaw depth is uniform, and deeper thaw appears to be associated with the water tracks. Enhanced thaw along water tracks is especially apparent in 1997, which was an abnormally wet year on the North Slope. The INV for the grid is high (average of 35%) and particularly pronounced at a number of the wetter grid nodes.
4.3.4. Sagwon Hills Flux Study Plot Sites
 Near the boundary between the Coastal Plain and Arctic Foothills physiographic provinces are two sites separated by about 4 km. These are 1-ha plots, so the results are not directly comparable to the 1000-m grids. Both sites are situated on gentle hillslopes, have similar aspects, and nearly identical macroclimate.
 Vegetation, however, is substantially different [Walker et al., 1998]. The moist acidic tundra site contains a thick (15 cm) organic layer, sphagnum moss, and relatively shallow average thaw depth (36 cm) owing to the effective insulating properties of the surface material [Bockheim et al., 1997, 1998; Hinkel et al., 2000]. By contrast, the moist nonacidic tundra site is dominated by grassy tussocks, a thin (9 cm) organic layer, frost scars, and a relatively thick active layer (55 cm). These sites are included in this discussion to demonstrate the influence of vegetation, especially sphagnum moss, on thaw depth. This effect results in a significantly different thaw response across relatively small distances.
4.4. Discussion of Foothills Sites
 In summary, sites in the Foothills demonstrate the influence of topography (Toolik Lake), vegetation (Sagwon sites), and small-scale variations such as water tracks (Imnavait Creek, Happy Valley) on thaw depth. These sites also show a high degree of node-specific interannual variability that makes predictive mapping of thaw depth problematic at the scale and resolution of the grids [Nelson et al., 1998a, 1999; Gomersall and Hinkel, 2001].
5. Temporal Patterns of Annual Grid Averages
 Grid averages for the 6-year period of record are shown in Figure 5. Consistent with Barrow's cold marine summer climate, average thaw depth is smallest at this site. Inland, Atqasuk experiences intermediate values. Both West Dock and Betty Pingo experience somewhat deeper thaw than the sites in the Foothills further to the south and at higher altitudes.
 Sites on the Coastal Plain experience relatively small average INVs, whereas those in the Foothills show substantially higher values. These results are consistent with those obtained by Nelson et al. [1998a, 1999] and Gomersall and Hinkel  from spatial autocorrelation, nested sampling, and geostatistical analyses. These studies demonstrated that most of the thaw variation occurs at short (1–10 m) sample separation distances at the Foothills sites, reflecting the influence of tussocks and frost boils on local heat flow. In contrast, most variation occurs at sample separation distances of 100–300 m at sites on the Coastal Plain, reflecting contrasts between the relatively dry “uplands” and wetter lake basins.
 The temporal patterns for the period 1995–2000 at the Foothills sites are shown in Figure 6a. Here, the dot represents the areally averaged annual thaw depth, with bars extending one standard deviation in either direction; triangles indicate the minimum and maximum thaw depths for that year.
 Mean thaw depths at the Foothills sites are relatively consistent over the latter half of the decade. The exception is 1998, which was one of the warmest and wettest years on record on the North Slope of Alaska and, indeed, in northwestern North America (Environment Canada, available at http://www.msc-smc.ec.gc.ca/ccrm/bulletin/annual00/, 1999). All sites experienced maximum average thaw in that year. By contrast, the summer of 2000 was cool, and minimum average thaw depths were recorded on all grids in that year. This demonstrates a regionally consistent response to air temperature forcing.
 The plots in Figure 6 use the same vertical scales. A cursory examination reveals that Coastal Plain sites have substantially greater dispersion around the mean value. Indeed, the average standard deviation for Coastal Plain sites, exclusive of Barrow, is nearly twice that of the Foothills sites. As discussed earlier, this reflects the influence of primary terrain elements (especially thaw lake basins) on the bimodal distribution. As demonstrated in Table 2, the standard deviation of the Coastal Plain uplands is similar to that for the Foothills sites. Drained lake basins have substantially deeper thaw and much larger standard deviations, which inflate the measurements of central tendency when applied to the entire grid.
 At Barrow, the record is somewhat longer. During the period 1962–1968, replicate measurements of thaw depth were made at nineteen 10 m × 10 m plots by researchers at the U.S. Army's Cold Regions Research and Engineering Laboratory (CRREL). These plots were distributed along a linear transect from Central Marsh to Elson Lagoon [Brown and Johnson, 1965], and were subsequently reoccupied in 1991 [Nelson et al., 1998b]. Thaw depths in the 1960s were relatively great and showed significant interannual variation (Figure 7a). No data are available for the period 1969–1990. The 1990s show a trend similar to that of the other CALM sites, but the pattern of gradual thaw deepening throughout the 1990s, peak thaw in 1998, and shallower thaw since that time, is more apparent. Thaw depth does not appear to be related to summer (June–August) precipitation.
 End-of-season thaw depth at each grid (Z) can be correlated with the accumulated degree days of thaw (ADDT: °C d) using a form of the Stefan solution given by:
where E is an “edaphic factor” with dimensions [m2/°C d)]1/2, representing soil thermal conductivity, density, moisture content, and latent heat effects [Nelson and Outcalt, 1987; Hinkel and Nicholas, 1995; Nelson et al., 1997; Klene et al., 2001]. ADDT is determined by using air temperature data, and summing the daily averages for the period beginning at thaw initiation and ending on the day of grid probing. Strong relations, as defined by a linear best fit and correlation coefficient, are apparent for most of the CALM sites (Figure 8) for the period of record, indicating that end-of-season thaw depth responds consistently to the local forcing function. Anomalously low thaw depth values in 2000 at Betty Pingo and West Dock strongly influence the correlation coefficient.
 In the case of the Barrow CRREL sites, two distinct clusters emerge when √ADDT is plotted against average annual thaw depth (Figure 7b). This reflects a fundamentally different response between the 1960s and the 1990s. Indeed, the same surface energy input in the 1990s produced only about 70% of the thaw penetration as that experienced in the 1960s, and implies a much higher soil thermal inertia during the more recent period. This also suggests that the active layer exhibits Markovian behavior such that the response function is “reset” following summers with extremely large or small degree-day accumulations [Nelson et al., 1998b]. Average maximum thaw depths in subsequent years cluster around a new level until resetting occurs following another extreme summer. This may be due to the growth of ice lenses at the base of the active layer or uppermost permafrost, which would have the effect of reducing the value of E. It is also possible that the Markovian shift might be triggered by extreme changes in soil moisture.
 The anomalous measurements collected at Betty Pingo and West Dock in 2000 may reflect another resetting event following the deep thaw in 1998 at these two coastal sites. The example from Barrow demonstrates the importance of obtaining long-term records of sufficient duration to detect such shifts. Collection of ancillary soil data will provide the means of identifying the triggering mechanism and more realistically model the active layer response.
 Six years of record permit several general conclusions: (1) Sites on the North Slope of Alaska respond consistently to forcing by air temperature on an interannual basis. All sites experienced maximum average thaw depth in 1998 and a minimum in 2000, consistent with the warmest and coolest summers during the period of record. End-of-season thaw depth is strongly correlated with local air or surface temperature over subdecadal timescales, but the active layer may exhibit Markovian behavior over longer durations. (2) There is significant intrasite variation in thaw depth and near-surface soil moisture content within each 1-km2 grid, reflecting the impact of vegetation, substrate, snow cover dynamics, and terrain. (3) On the Coastal Plain, average thaw depth and thaw depth variation is significantly greater in drained thaw-lake basins, where soils are typically at or near saturation. This results in a bimodal distribution of thaw depths related to primary landscape elements. (4) Foothill sites demonstrate large spatial and node-specific interannual variability resulting from microtopography and temporal fluctuations of soil moisture content; this makes predictive mapping of thaw depth problematic at the scale and resolution of the grids. The spatial pattern of thaw depth across sites on the Coastal Plain is relatively consistent in the uplands. Thaw-lake basins and lake margins, however, exhibit more complex patterns attributable to fluctuating water levels.
 This research was supported by grants from the National Science Foundation, Office of Polar Programs, to K.M.H. (OPP-9529783, 9732051, and 0094769) and F.E.N. (OPP-9612647, 9907534, and 0095088). We are grateful to the Barrow Arctic Science Consortium for administrative assistance and to the Ukpeagvik Inupiat Corporation for access to the Barrow Environmental Observatory. We thank J.D. Fagan, C. Gomersall, A.E. Klene, I. Maximov, L.L. Miller, G.R. Mueller, and N.I. Shiklomanov for their efforts in the field and laboratory over the 6-year period. The ARCSS/CALM grids were conceived and implemented in the 1980s and early 1990s by L.D. Hinzman, D.L. Kane, D.A. Walker, J. Brown, and the late K.R. Everett under several programs funded by the U.S. Department of Energy and the National Science Foundation.