We evaluated the development of lake basins on the central coastal plain of northern Alaska on the basis of topographic profiles, soil and ground ice surveys, radiocarbon dating, photogrammetric analysis, and regional comparisons. Our analysis reveals that lake evolution is much more complex and less cyclic than theorized by previous investigations. In the area we studied, there was insufficient ground ice in the oldest terrain to form thaw lakes, the aggradation of ice in the margins of drained lake basins was insufficient to heave the surface up to near original topographic conditions, and the process occurs at too slow a rate for a “thaw lake cycle” to develop within the Holocene. Accordingly, we revised the conceptual model of lake and basin development to be consistent with the patterns and process we observed on the extensive sand sheets underlying most of the coastal plain. Developmental stages include (1) initial flooding of depressions to form primary lakes, (2) lateral erosion, with sorting and redistribution of sediments, (3) lake drainage as the stream network expands, (4) differential ice aggradation in silty centers and sandy margins, (5) formation of secondary thaw lakes in the heaved centers of ice-rich basins and infilling of ponds along the low margins, and (6) basin stabilization.
 Numerous variants of a thaw lake cycle have been proposed, but most lack complete descriptions of the processes by which the surface returns to original conditions (i.e., a return of the surface form to that present prior to the onset of the cycle). A cycle was first proposed by Cabot  on the basis of interpretation of lake patterns evident on aerial photographs; this concept emphasized thaw lake formation and drainage and omitted ice aggradation in drained basins. Britton  articulated a more complete thaw lake cycle that involved (1) initial flooding of basins, (2) lake expansion and coalescence through lateral mechanical erosion and thawing accompanied by material sorting, (3) drainage, (4) ice wedge development in drained basins, and (5) secondary development of thaw ponds from ice wedge degradation. This concept provided little detail on the ice aggradation process and does not complete the cycle by recreating the original upland surface. Carson  described a lacustrine cycle that included (1) initial development of thaw ponds, (2) expansion and deepening of ponds through lateral erosion and subsurface thawing in a youthful stage, (3) basin elongation perpendicular to the wind due to longshore currents and thawing at the ends in a north-south direction during the mature stage, (4) drainage by stream migration, and (5) a secondary cycle of ponds on shelves behind barrier beaches and strands with persistence of the deeper central lake at the old stage. The role of ice aggradation and the processes involved in return to the original surface are not included in his cycle. Everett  defined a more complete cycle beginning with an ice-rich raised surface that included (1) climate change or surface disturbance that initiates permafrost degradation, (2) degradation of ice wedges and development of small thaw ponds, (3) expansion of the thaw pond by surface and subsurface thawing, (4) expansion into large lakes by bank erosion and subsurface thawing accompanied by material sorting, (5) partial or complete drainage by stream capture or breaching, and (6) and reestablishment of ice wedges and surface polygon patterns. Finally, Billings and Peterson  described a thaw lake cycle with 10 stages that is similar to Everett's, but emphasizes the role of ice wedge aggradation and degradation within basins.
 Although similar, the concepts vary in their explanations of lake formation, in the roles attributed to ice aggradation and degradation, and in their treatments of the return of surfaces to near-original conditions to complete the cycle. They all assumed that initial conditions had substantial ground ice and were favorable for thermokarst, an assumption we believe to be problematic. Most authors recognized that sediments are sorted and redistributed during lake expansion, yet the importance of this redistribution to ground ice dynamics has not been recognized. While attention has focused on ice wedges, the nature and distribution of other types of ground ice have been overlooked. Finally, previous investigators have hypothesized that the surface returns to near-original conditions, thus creating a cycle, but the lack of stratigraphic evidence of distinctive lacustrine deposits [Kidd, 1990; Murton, 1996] in raised surfaces outside of lake basins, indicates that much of the terrain has not been affected by a complete cycle. Consequently, the primary reason a consensus has not emerged on the concepts underlying the thaw lake cycle is that these studies were not supported by quantitative data on topographic changes, soil stratigraphy, and ice volumes necessary to evaluate the physical processes associated with thaw lake development.
 In this paper, we reevaluate the concept of the thaw lake cycle, based on detailed terrain analysis, field surveys, and photogrammetric analyses. First, we briefly summarize information from studies [Jorgenson et al., 2003, 2005] we conducted on terrain, permafrost, and landscape change near Fish Creek on the central Beaufort Coastal Plain in northern Alaska (Figure 1) to evaluate the assumptions of existing thaw lake conceptual models. Our analysis focuses on (1) whether enough ground ice is present to allow the development of thaw lakes, (2) whether ice aggradation after lake drainage is sufficient to facilitate return to original conditions, and (3) whether the process occurs at a rate sufficient to allow the cyclic development of thaw lakes. Second, we develop a new conceptual model of the evolution of lakes and drained basins based on our field and photogrammetric analyses. Finally, we present a map of the distribution of thaw, depression (nonthaw), riverine, and delta lakes on the coastal plain based on terrain relationships and landscape-scale mapping. While orientation is a prominent characteristic of lakes on the Beaufort Coastal Plain [Cabot, 1947; Black and Barksdale, 1949; Carson and Hussey, 1962; Sellman et al., 1975; Carson, 2001], it not central to an understanding of the evolution of lakes and drained basins across the coastal plain landscape and is not evaluated in this paper.
2. Evaluation of Thaw Lake Models
2.1. Ground Ice Volume and Thermokarst Potential
 In the classical concept of thaw lake development, formation of water bodies begins with degradation of ice wedges in ice-rich upland terrain [Shumskiy and Vturin, 1966; Tomirdiaro and Ryabchun, 1973; Everett, 1980]. Degradation is most intense at the intersections of ice wedges, and deepening of the water in the troughs leads to the formation of small, deep (>1.5 m) ponds. A thaw bulb then develops under the deep water and the thaw lake expands laterally through both mechanical and thermal erosion.
 The amount of ground ice and the potential for thermokarst to create thaw lakes in our study area varied greatly among the geomorphic units across the landscape (Figures 2 and 3) . Alluvial marine deposits (21% of study area) are the oldest deposits to stabilize at the end of the Pleistocene and are composed of massive, slightly pebbly, silty sand of problematic genesis, capped with a thin layer of eolian silt [Jorgenson et al., 2003]. Eolian inactive sand (7%) is widespread from Fish Creek to the Meade River and is composed of cross-bedded very fine sand. Ice-rich, drained basin margins (26%) have a thick surface organic layer underlain by stratified fine sands and are characterized by the presence of distinct low-centered polygons formed from ice wedge development. Ice-rich, drained basin centers (11%) have thick deposits of nonstratified ice-rich silts, distinct low-centered and high-centered polygons, and usually are raised higher than the adjacent margins. Ice-poor, drained basin margins (2%) and ice-poor, drained basin centers (1%) are similar to their ice-rich counterparts, but lack visible ice wedge development. Meander inactive overbank deposits (6%) are composed of interbedded organic and silt layers. Meander abandoned overbank deposits (1%) have a thick surface organic horizon over interbedded organic and silt layers and have high densities of low-centered and high-centered polygons. Deep isolated lakes (9%) and shallow isolated lakes (3%) are the predominant types of water bodies.
 When comparing the volumes of ground ice among geomorphic units, Jorgenson et al.  found that the mean ice volumes of segregated and pore ice in samples from permafrost within the top 2 m of the ground were highest in alluvial marine deposits (71%) and ice-rich, drained basin centers (66%), intermediate in ice-rich, drained basin margins (62%) and ice-poor, drained basin centers (59%), and lowest in ice-poor, drained basin margins (48%) and eolian inactive sand (45%) (Figure 4). That portion of the ice that was estimated to be in excess of soil pore space, ranged from 6% in eolian inactive sand to 43% in meander abandoned overbank deposits. Variations in the mean volume of ice wedges in the top 2 m were even greater, ranging from <1% in the ice-poor margins of drained lake basins, to 22% in delta abandoned overbank deposits, to an unusually high value of 33% in the ice-rich centers of lake basins, as calculated from estimates of the cumulative lengths of ice wedge troughs per hectare obtained from aerial photographs and ice wedge cross-sectional dimensions determined from bank exposures [Jorgenson et al., 2005].
 Ground ice volumes decrease rapidly with depth in geomorphic units with thin layers of surface material over the underlying sand sheet (ice-poor, and ice-rich, drained basin margins and eolian inactive sand). In other geomorphic units, ice volumes are high throughout the top 1.5–2 m of the profile. In deeper coring, Lawson  found ice-rich sediments generally were limited to the upper 3–5 m of coastal plain deposits in the Fish Creek area, indicating the lack of ice-rich deposits of Pleistocene age on the coastal plain. In addition, he found very ice-rich soils (>80% ice) were found to depths of 5 m in the center mounds in old basins and in upland silt in the foothills, where massive ice extended to depths of 22 m.
 These ground ice contents indicate that most terrain in the central Beaufort Coastal Plain does not have sufficient ice to allow the development of thaw lakes, except for the ice-rich centers of drained basins. The organic-rich silts deposited in the centers are highly susceptible to ice segregation and are easily deformed by ice wedge growth. In addition, the formation of ice at greater depths during downward freezing into the talik underneath the newly exposed sediments probably contributes more ice that was too deep (>3 m) to core in our field studies. Of particular interest to thaw lake formation, is the low to moderate ice contents in the eolian inactive sand and alluvial marine deposits, respectively, which comprise the dominant terrain units of early Holocene age in our study area in which the lake basins formed. Thermokarst in these units is limited to formation of deep polygon troughs and high-centered polygons [Jorgenson et al., 2006]. We did not observe any small, newly developing thaw lakes in these widespread deposits. Even severe disturbances in the region, such as the peat roads bladed during early oil exploration, have created only discontinuous deep thermokarst pits and high-centered polygons, not deep thaw ponds.
2.2. Aggradation of Ground Ice Toward Original Conditions
 Aggradation of ice in drained lake basins is central to the concept of a thaw lake cycle. We observed large differences in ground ice and relative elevations among the drained basin deposits. Along the sandy margins of lakes, permafrost aggradation is limited to the active layer that forms under shallow water. Under these conditions, the layered and reticulate ice that forms as permafrost aggrades can be abundant, but is relatively thin. Consequently, the ice aggradation is insufficient to heave the surface up more that a few decimeters (Figure 3). Similarly, the development of ice wedges is substantially restricted in the sandy material due to the higher resistance to deformation of the frozen sands and to the low thermal contraction coefficients of the sands with relatively low pore space volume. As a consequence of the restricted aggradation of ice in the sandy margins and low frost heaving, the margins are long-lasting, low features of the landscape (Figures 2 and 3).
 In contrast, the organic-rich silty centers are much more prone to ice aggradation due to the high frost susceptibility of the materials, downward freezing of ice into the saturated unfrozen sediments in the talik (unfrozen zone between active layer and underlying permafrost), and the rapid development of ice wedges in the easily deformed sediments. These deposits had the highest volumes of segregated and wedge ice in the top 2 m of any geomorphic unit (Figure 4). Furthermore, we suspect ice content is even higher at greater depths in the deposits where layered ice would have developed in the refreezing talik. The elevations of these deposits relative to the adjacent margins indicate that the surface typically heaves 2-4 m during refreezing (Figure 3). These are the only deposits in which the volume of ice is sufficient for development of thaw lakes. The low amount of ice aggradation in the sandy margins of the drained basins indicates that the basins become permanent features and do not heave back to conditions similar to the original eroded terrain. In areas of thermokarst lakes, such as near Barrow, high-resolution IFSAR topographic data reveal that elevations in old basins can be similar to the surrounding terrain, indicating that in areas with more uniformly distributed silty sediments, ice aggradation raises the surface throughout old basins to near original conditions.
2.3. Rates of Landscape Change
2.3.1. Surface Stabilization
 Dates from the base of the peat provide a good indicator of when erosional and deposition processes ended and the surface became stabilized by vegetation establishment. Dates for basal peat are the oldest in alluvial marine deposits (calibrated age range 9580–16,225 years B.P., n = 3) and eolian inactive sands (8810 years B.P., n = 1) as reported by Jorgenson et al. . In the ice-rich, drained basin centers, basal peat dates were intermediate (5075–8315 years B.P., n = 2). In ice-poor, drained basin centers, one core was sampled and yielded a basal carbon date of 10,310 years B.P. at 195 cm. Dates from ice-rich, drained basin margins were younger than other terrain types (840–5435 years B.P.). Hinkel et al.  also reported 14C ages of 2075–5370 years B.P. for the oldest drained lake basins near Barrow, Alaska.
2.3.2. Shoreline Erosion
 Shoreline erosion and water body contraction in three small areas (1450 ha) during the 46–56 year period from 1945–1955 to 2001 were evaluated using photogrammetric analysis by Jorgenson et al.  (one area in Figure 5). The analysis revealed shoreline erosion rates are greatest in large, deep lakes. These lakes are more abundant across the landscape and covered a mean of 26.6% of the three study areas (range 14.6–40.4%). Shallow lakes covered a mean of 6.2% (range 5.0–7.8%) and typically were small (<5 ha). Overall, 0.7% (range 0.4–0.9%) of the land in the three study areas was lost to shoreline erosion. Erosion was most prevalent in ice-rich, drained basin deposits. The maximum observed erosion rate was 0.8 m/yr, but in most areas erosion rates were much lower. The average annual erosion rate for the three study areas ranged from 0.008–0.017 %/yr of total land and water area (mean = 0.013 %/yr). Another striking feature of the change analysis was the decrease in size of many small, shallow lakes, which was attributed to high water levels in 1945 and to difficulties in interpreting shorelines due to the poor quality of the 1945 photography.
 Both the old ages of the basins and the slow rates of lake erosion indicate that the geomorphic processes are too slow to allow development of a thaw lake cycle during the Holocene when the surficial deposits became stabilized. The common age of the old drained basins, dating from the mid to late Holocene, indicates that the drained basins are long-persisting features and are not easily modified by lacustrine processes. The observed erosion rates also do not appear to be sufficient to allow development of a thaw lake cycle, based on the assumption that the rate of erosion was constant over both time and space during the Holocene. Using the recent rates of erosion, and assuming the lakes eroded out from a central point, the water bodies would have required 2466 years (range 1995–2875 years) to reach their present sizes. Although this simplistic approach ignores numerous factors affecting erosion rates, it indicates that lake formation, drainage, and ice aggradation are too slow to have allowed cyclic thaw lake formation on surfaces that stabilized during the early Holocene.
3. New Model of Evolution of Lakes and Drained Basins
 Because the concept of a thaw lake cycle is not supported by evidence from our study area, we developed a new conceptual model of lake and drained basin development that incorporates the early stages of some of the older conceptual models and revises the older stages. The revised model includes six main stages of development (Figure 6): (1) initial flooding of primary lakes, (2) lateral expansion, with sorting and redistribution of lacustrine sediments, (3) lake drainage, (4) differential ice aggradation in silty centers and sandy margins, (5) secondary development of thaw lakes in the ice-rich silty centers and infilling ponds along the ice-poor margins, and (6) lake stabilization. This conceptual model applies to sand and slightly pebbly silty sand deposits that cover most of the Beaufort Coastal Plain. In the following discussion of the various developmental stages, we evaluate the stages in terms of topographic changes, soil stratigraphy, ice content, and relative ages.
3.1. Initial Flooding of Low-Lying Depressions
 In our interpretation, lakes over most of the coastal plain were formed initially by flooding of depressions within the undulating ground surface during the beginning of the Holocene, when the surface stabilized and climate ameliorated from the cold and dry conditions of the late Pleistocene [Carter et al., 1984; Rawlinson, 1983, 1993]. Vegetation growth and peat accumulation stabilized the ground surface by about 8000–11,000 years ago [Carter et al., 1984; Jorgenson et al., 2003]. Dating of basal peat from the oldest upland surfaces (alluvial marine, old alluvial terrace, and eolian inactive sand deposits) yielded calibrated ages (2 sigma error range) of 9460–9700, 10,260–10,690, and 15,870–16,580 years B.P. [Jorgenson et al., 2003]. Coincident with this surface stabilization was the initiation of lakes 9000 to 10,000 years B.P. [Ritchie et al., 1983; Rawlinson, 1993]. Similarly, the oldest date for basal organic material from the centers of deep lakes in our studies was a calibrated calendar date of 8200–8430 years B.P. [Jorgenson et al., 2003]. The lack of older lakes is attributed to cold, arid, and windy conditions and active eolian deposition near the end of the Pleistocene Era [Carter et al., 1987; Rawlinson, 1993].
 During the period when the sand sheets were stabilizing at the end of the Pleistocene, we believe the ground ice content was too low for development of thermokarst lakes for several reasons. First, nearly all the surficial deposits are Holocene to late Pleistocene age and the newly stabilized deposits did not have sufficient time to aggrade ice deep in the soil profile. Second, we are not aware of any reports of deep massive ground ice on the coastal plain, other than that related to Holocene ice wedge development. None of the stratigraphic descriptions of mine site exposures in the central Beaufort Coastal Plain by Rawlinson , deep coring (10–30 m) associated with oil exploration in the NPRA [Lawson, 1983] and for the Trans-Alaska Pipeline [Kreig and Reger, 1982], and from our recent observations of coastal bluffs at 50 sites along the Alaskan Beaufort Sea Coast revealed deep Pleistocene sediments with massive ice. There are extremely ice-rich soils of late Pleistocene age in the loess belt along the lower foothills [Carter, 1988], however. Consequently, it is highly unlikely the first lakes to develop on the coastal plain during the early Holocene were formed by degradation of ground ice. Rather, these depression lakes formed simply by the accumulation of water in low-lying areas across the undulating surface and many of the deep lakes that exist today simply are remnants of these old primary lakes. This mode of development of the first lakes is consistent with the initial depression flooding in some previous models [Britton, 1957; Shur, 1977; Gravis, 1978].
3.2. Lateral Expansion and Sediment Redistribution
 After initial flooding, lake levels would have fluctuated in response to changes in precipitation and other components of the water balance, and shorelines would have expanded as a result of wave erosion. In our studies, we observed that the older alluvial marine deposits were eroding principally through mechanical erosion, leaving characteristic wave-cut benches. Typically, the eroding banks are 1.5–2 m high and water depths increase slowly from <0.1 m near the bank to 0.3–0.5 m at a distance of tens of meters. In contrast, thaw lakes in ice-rich terrain, such as the Colville River Delta, have steep shore profiles, have banks that are undercut by a thermal niche, and reach water depths of 3–4 m within 10 m of the banks [Jorgenson et al., 1997].
 Redistribution of sediments during early lake formation and expansion, with concomitant accumulation of fines and organics in the deepest portions of the lakes, is fundamental to the dynamics of later ice aggradation and degradation. Numerous investigators have observed the prominence of sandy sediments along the margins of large lakes and the accumulation of fine-grained sediments and organic material in the centers [Britton, 1957; Carson and Hussey, 1962; Carson, 1968; Tedrow, 1969; Murton, 1996]. In our studies, sediments near the margins of ice-rich, drained basins were dominated by massive and layered sands (fine and medium sands), whereas the sediments in the centers of ice-rich, drained basins were dominated by massive fines with organics, turbated fines with organics, and limnic fines (algal rich). Although small, buried peat mats are common along the sandy margin, fibrous mats and organic-rich debris are more abundant in retrogressive thaw slumps and debris flows associated with thermokarst lakes in the Tuktoyaktuk Coastlands in northwestern Canada [Murton, 1996]. These differences in particle size from sandy margins to silty centers in turn control the differential development of segregated and wedge ice. Shoreline erosion, lake expansion, and redistribution of sediments were partially recognized in the thaw lake concept of Britton .
3.3. Lake Drainage
 The partial or complete drainage of lakes by stream capture, shoreline breaching, or coastal erosion is a dramatic and frequently observed phenomenon in some regions of the coastal plain [Cabot, 1947; Hopkins, 1949; Britton, 1957; Carson and Hussey, 1962; Everett, 1980; Mackay, 1988]. While the process of lake drainage appears straightforward, differences in the extent and rate of drainage result in a wide range of subsequent basin water levels and surface conditions. In addition, lake levels can fluctuate over time due to climatic changes [Are, 1969; Bosikov, 1988], or be drawn down slowly as the outlet channels progressively erode and lower the base levels of the lakes. The multiple processes causing water level changes thus create a range of surface ages from initial exposure. This stage of basin development is incorporated in all previous conceptual models, although substantial uncertainty remains about the relative importance of tapping versus climatic change in causing lowered water levels in the lake basins.
 The radiocarbon dating of basal peat from drained basins at Barrow [Carson, 1968, 2001; Hinkel et al., 2003] and the northeastern part of the National Petroleum Reserve–Alaska (NPRA) [Jorgenson et al., 2003] indicates that lake drainage was most active 1000–5000 years B.P. The lack (<0.01% of area) of recently drained lakes with barren sediments on the coastal plain east and west of the Colville River [Jorgenson et al., 1997, 2003] also indicates that drainage was more active in the past. Given that the early stage of development is rare and the old stage of development is abundant, the distribution of stages does not appear to be sufficient to support an ongoing cycle. Recently drained lakes are more prevalent in ice-rich regions, such as near the coast at Barrow [Sellman et al., 1975] and in the Colville River Delta [Jorgenson et al., 1997], presumably due to faster shoreline erosion or channel migration.
 The ages of most drained lake basins indicate that it took considerable time from when the surface stabilized in the early Holocene to develop a drainage network with sufficient integration to tap the isolated lakes. Once drainage has occurred in the initial depression lakes, however, further drainage of the remaining lake is uncommon, because the elevations of the resulting water surface and the bottoms of lakes are much lower than the outlet. Further draining of the isolated lakes would require much deeper incision of channels across the landscape and gradients are too low for such scouring.
 The extent of early primary lakes before drainage can be estimated from the distribution of lacustrine deposits. Terrain unit mapping by Jorgenson et al.  indicates that all drained basin deposits combined cover 39.4%, and freshwater lakes (deep and shallow lakes combined) currently cover 15.3%, of the mapped area. Taken together they indicate that primary lakes once covered or affected 54.7% of the area. Currently, deep open lakes cover 8.9% of the mapped area. We believe that most of the deep lakes are the remnants of the central portion of the larger primary lakes after partial drainage.
3.4. Differential Ice Aggradation
 After drainage, the nature and volume of ground ice that develops in the newly exposed sediments are highly variable across the basins, depending on the texture of the redistributed sediments and on thaw bulb development in the former lakes where water is greater than 1.5–2 m deep [Brewer, 1959]. The general trend, however, has been for little ice aggradation in the sandy sediments along the margins, where thaw bulbs do not develop, and abundant ice aggradation in the organic-rich centers where thaw bulbs typically develop under deep water.
 Differences in sediment textures between the organic-rich silts in the deep centers of the lakes and the sandy sediments along the shallow margins provide the conditions for widely varying patterns and processes of ice development across the drained basins. The dynamics of the original active layer along the margins, and development of a thaw bulb in the centers of deep lakes, also greatly affect ice development. Refreezing within the center of the drained basin creates substantial heaving that raises the surface. Most of the uplift probably occurs within the first 100 years following drainage. In addition to this uplift from layered ice formed at the downward freezing front in the underlying thaw bulb, segregated ice (organic matrix, ataxitic, and reticulate, described by Shur and Jorgenson ) formed during upward freezing associated with readjustment of the active layer [Shur, 1988], and epigenetic wedge ice formed by thermal contraction at the surface, also contribute to the volume expansion. Consequently, the tops of the ice-rich centers can rise 1.5–4 m above the bottoms of the ponds along the margins of the basin (Figure 3).
 Under certain circumstances, pingo ice can form during freezing of water in sediments within the closed talik that forms under the deep-water zone in a lake. Although we observed a few small pingos, we have not included pingos in our simplified conceptual model of lake basin evolution because of their rare occurrence in the studied area. Similarly, we have observed the frequent occurrence of small ice cored mounds (1–5 m diameter) in younger ice-poor basins that form during freezing of suprapermafrost water at the base of the active layer, but these features contribute only negligible amounts of ice to the overall deposits.
 Along the margins of drained basins, ice aggradation is affected by the thickness of the original active layer below the shallow margins of the lake, the sandy sediments, and surface organic accumulation. Where water is shallower than 1.5–1.8 m, pond sediments have a moderately thick (0.6–0.8 m) active layer that is important because it constrains the volume of material in which new segregated ice can develop by upward freezing. After drainage, initial freezing and readjustment of the active layer occurs within a winter or two. Thereafter, the active layer slowly thins as the thermal regime readjusts to increased vegetation development and organic accumulation. During this readjustment, thin layers of ataxitic and reticulate ice are prevalent, but the volume of segregated ice is limited by the thickness of the original active layer under the lakes, and the presence of sandy sediments near the surface. The accumulation of organic material at the surface adds to the soil volume and affects the structure and volume of ice that develops. In our studies, the cumulative thickness of organics near the surface averaged 0.66 m in the ice-rich margins of basins. The underlying massive and layered sand, however, had predominantly pore ice with a mean ice volume of 44% [Jorgenson et al., 2003]. In long-term monitoring of an experimental drained lake in northwest Canada, Mackay and Burn [2002a] found that aggradation of ice in the thinning active layer contributed ∼10 cm of heave over 20 years. Our limited surveys revealed that the older ice-rich margins of basins were only 0.7 to 1.5 m higher than the bottoms of shallow ponds in the more recently exposed ice-poor portions of the same drained basins (Figure 3). Because the heaving of the soils in these sandy margins is low, they persist as areas of low-lying and usually flooded terrain.
 Development of ice wedges in newly exposed sediments, or under shallow water that freezes to the bottom during winter, makes an important contribution to ice volume and greatly affects surface water redistribution. Development of the wedges follows a progression of microtopographic changes that includes (1) nonpatterned ground, (2) disjunct low rims, (3) low-density, low-centered polygons, (4) high-density, low-centered polygons, and finally (5) mixed high- and low-centered polygons. While initial ice wedge growth can be as high as 3 cm/yr, growth slows rapidly as vegetation and increased snow entrapment reduce winter temperature fluctuations [Mackay and Burn, 2002b]. Analyses of floodplain development on the Colville River indicates that disjunct rims take ∼300–500 years to develop, low-density, low-centered polygons take ∼500–1500 years, and complete development of high-density, low-centered polygons can take 1500–3000 years [Jorgenson et al., 1998]. At the last stage, the ice wedges typically are 2–3 m across the top and occupy ∼20–30% of the volume of the top 2 m of soil (Figure 4). Development of this wedge ice deforms the adjacent sediments and contributes to heaving of the land surface. While wedge ice can develop in both the sandy margins and organic-rich centers, they achieve larger volumes in the centers because of the more easily deformed organic-rich sediments.
3.5. Secondary Development of Lakes Within Basins
 Although the secondary development of water bodies within the ice-rich, drained basins is complicated by large variations in sediment and ground ice, we have identified two main types of water bodies that can reoccupy the drained basins: (1) small, shallow infilling ponds caused by impoundment of water in the low-lying margins of the basin; and (2) deeper, larger lakes created by thawing of the organic- and ice-rich materials in the center. These differences, established by the evaluation of sediment characteristics, photogrammetric analysis of lakeshore erosion, and photointerpretation of water body characteristics, are discussed below.
 We attribute the formation of numerous small, shallow ponds around the margins of the drained basins primarily to impoundment of water in the lowest portions of the basins. Because of ice aggradation in the organic-rich silts and heaving of the basin centers, the margins of the basins become the lowest portion of the landscape and therefore collect water. Initially, the water that collects in these low-lying areas can form large ponds, but over time accumulation of organic matter and ice aggradation in the adjacent tundra cause the ponds to become subdivided and reduced in size. During this stage, organic material is added both as algal material in diatomaceous benthic mats within the shallow ponds and as fibrous sedge peat in the wet tundra surrounding to the ponds. In later stages of development, these ponds become shallow because of thick accumulations of diatomaceous benthic algae at the bottom, and tend to have round to polygonized and abrupt vegetated margins.
 Evidence for pond paludification, or infilling, is provided by soil stratigraphy and photogrammetric analysis of 1945 and 2001 imagery. The stratigraphy of adjacent soils frequently revealed buried layers of well-preserved, olive green limnic material, indicating that the ponds have been infilling and shrinking. The thickness of the tundra organic mat (mean = 0.66 m in ice-rich basins) is similar to the elevation difference between pond bottom and adjacent wet tundra surface (Figure 3). This organic matter helps elevate the surface, creates surface conditions favorable for plant growth and development of wet and moist tundra vegetation, and contributes to the abrupt peat margins along most small ponds. Continued accumulation of both benthic algal mats and sedge peat helps transform the ponds from aquatic sedge marsh along the margins to wet sedge meadow tundra and causes the pond area to shrink. During the accumulation of this organic matter, the depth of the active layer decreases so that some of this organic material becomes incorporated into the permafrost.
 Reinforcing this process is aggradation of both segregated and wedge ice. Ice volumes along the margins of ice-rich, drained basins with well-developed ice wedges are highly variable (ranging from 40 to 80%) in the top 2 m, but decrease to 40–50% below 2 m. At an intermediate stage of ice wedge polygon development, wedge ice associated with low-density, low-centered polygons contribute ∼10% to the volume of materials near the surface. Limited data from Jorgenson et al.  indicates surface accretion (through organic matter and ice accumulation) averages 0.4 m/1000 yr and therefore is sufficiently rapid to help modify the shoreline configuration. Photogrammetric analysis of 1945 and 2001 photography indicates that pond margins are remarkably stable and that erosion is minimal (Figure 5). The lack of evidence for thermokarst processes provides further support for paludification. If the ponds were of thermokarst or erosional origin, they would have measurable rates of retreat (particularly if due to thermokarst), organic mat fragments would be scattered across sandy bottom sediments from erosion, and algal remains would not be present in the adjacent soils.
 In contrast to the shallow, infilling ponds formed through basin impoundment and paludification, some of the deep lakes in the centers of the basins are true thaw lakes formed from degradation of ice-rich materials in the centers. In some situations, ponds developed near ice-rich centers can cause lateral thermokarst erosion of basin centers. Evidence for thermokarst includes: (1) abundant excess ice in degrading soils at the eroding front; (2) an abrupt face (0.7–1.0 m) and underlying thermal niche where the shoreline is thawing; (3) organic mats slumping over the eroding shoreline indicating that material is being lost by thawing of ice rather than mechanical erosion of organic and inorganic material; (4) abundant fragments of organic mats strewn across sandy lake sediments; and (5) scalloped shorelines created by more rapid degradation of wedge ice, leaving the rounded polygon centers protruding into the lake.
 These true thaw lakes are constrained to the centers of old basins. Secondary thaw lakes resemble primary depression lakes, because both can be large with deep water in the centers. Remnant primary lakes typically have shorelines with multiple fringes of beach ridges and old shorelines that indicate drainage rather than expansion, or have wide sandy wave-cut benches with gently sloping bottoms.
 Formation of new thaw lakes developing in the original deep portion of primary lakes (ice-rich centers) is part of a one-way trajectory involving initial flooding in low-lying basins, sediment sorting, drainage, and formation of extremely ice-rich materials. These are necessary precursors to the development of a thaw lake in areas with sandy deposits. Once lakes form from thermokarst in the centers, however, surface water elevations are low relative to the surrounding older surfaces. Future drainage and continuation of the cycle are unlikely. This is even more important in areas with deep thermokarst basins in extremely ice-rich materials, such as the thick silt deposits in the lower Brooks Foothills and the Tuktoyaktuk Coastlands. This pattern is different from the traditional concept where old, higher terrain outside the basins degrades to form lakes, the lakes drain, and then return to an elevated surface by ice aggradation in a continual reworking of the landscape.
3.6. Basin Stabilization
 In the final stage of basin development, the shallow, infilling ponds and deeper thermokarst lakes that develop in the former ice-rich centers become persistent features (Figure 6). The infilling ponds are too small, and the shorelines have too much thick fibrous peat, to be susceptible to wave erosion. They continue to accrue diatomaceous benthic material in the bottoms and peat along the shores. While the surface around the ponds can be susceptible to minor ice wedge degradation, thermokarst is insufficient to initiate large ponds because of the low thaw settlement properties of the organic and sandy material in the basin margins [Jorgenson et al., 2006]. The abundance of old, indistinct basins with numerous small, rounded ponds indicates that this stage is stable (Figure 2). The ponds usually become further subdivided as ice wedges develop under the shallow water and the polygon rims protrude above the water. This process has often been mistaken as part of early stages of thermokarst development and coalescence of small, expanding thermokarst ponds.
 The deep thermokarst lakes that develop in the basins centers are persistent features. Because lakes have reoccupied the lower centers, and the sandy margins have developed thick organic accumulations with low thaw settlement properties, the possibility of further tapping and drainage is low. Although basins adjacent to large, meandering rivers are susceptible to tapping and complete drainage, most thermokarst lakes in the basin centers away from larger floodplains have little likelihood of further lowering because stream gradients are too low.
 The ice-rich centers of basins do not all degrade into ponds, however. Many basins have persistent, domed centers surrounded by small, shallow, infilling ponds. At the oldest recognizable stage, the domed centers become indistinct and the infilling ponds are reduced to tiny round remnants. While some ice-rich centers have degrading ice wedges, the resulting high-centered polygons appear to have sufficient peat accumulation to resist degradation into larger ponds.
4. Distribution of Lake Types
 Using this conceptual model of lake-basin formation and the traditional concept of thaw lakes, we developed a map of lake types across northern Alaska based on lake morphology and association of lakes with surficial materials of varying ice contents (Figure 1). We differentiated four types of lakes.
 1. True thaw lakes are associated with silty marine, deltaic, and loess deposits that are extremely ice-rich. The lakes tend to be either round or elliptical in shape.
 2. Basin (nonthaw) lakes are formed in depressions in undulating, sandy alluvial marine and eolian deposits with low to moderate ice contents, as described above. The lakes are rounded to elliptical, but orientation is parallel to the intervening broad ridges.
 3. Riverine lakes are associated with abandoned or inactive river channels.
 4. Delta lakes include thaw lakes in the upper part of the delta, riverine lakes in abandoned channels, and tidal lakes in depressions in the lower part of the delta.
 Thaw lakes are found predominantly in areas with thick silt deposits (Figure 7a), such as the narrow coastal plain extending from Barrow to Cape Halkett, which is underlain by marine silts and clays and the Colville River Delta [Jorgenson et al., 1998]. The marine silts near Barrow can have volumetric ice contents >80% [Brown, 1968]. Small thaw lakes occur on the distal, higher portions of abandoned floodplains with thick, silty overbank deposits [Jorgenson et al., 2003] and are particularly abundant near the lower Ikpikpuk River. Deep thaw lakes formed in massive ice in Late Pleistocene loess deposits (Figure 7b) are common in the lower Brooks Foothills [Carter, 1988]. In addition, extensive areas of thermokarst occur on the Seward Peninsula [Hopkins, 1949; Kidd, 1990] and Siberian Lowlands, which have thick accumulations of Pleistocene massive ice.
 Basin lakes are dominant on most of the Beaufort Coastal Plain, including most of the NPRA, Prudhoe Bay and Kuparuk region, and the Arctic National Wildlife Refuge (ANWR). The western part of the region is covered by a large eolian sand sea [Carter, 1981] and most of the coastal plain from northeastern NPRA to ANWR is covered by slightly pebbly silty sand that has variously been termed alluvial marine and alluvial plain deposits [Carter and Galloway, 1985], eolian sands(Beechey Sands of Rawlinson ), and Quaternary silt and sand over fluvial gravel and gravelly sand [Carter et al., 1986]. The sand dune deposits usually contain only pore ice (Figure 7b), while slightly pebbly silty sand of alluvial marine deposits (Figure 7c) have higher silt contents, thicker surface organic deposits, and moderately high ice contents. The landscapes dominated by basin lakes also have a minor amount of secondary thaw lakes that are eroding into the ice-rich centers of drained basins.
 Riverine lakes are common along large meandering river floodplains on the coastal plain, and to a lesser extent on braided floodplains in the foothills. The lakes are oriented along the sinuous channels of the riverbed and shapes are not affected by wind erosion. Areas mapped as being dominated by riverine lakes also have a minor number of thaw lakes developed on narrow abandoned floodplain deposits.
 Delta lakes are found on all the deltas along the Beaufort Sea Coast. They are most abundant on the Colville Delta. Lakes range from freshwater in the upper deltas to saline along the outer tidal flats. Thaw lakes on the Colville and Ikpikpuk River deltas, which are mostly <2000 years old, are oriented NE-SW parallel to the winds.
 The concept that thaw lakes continually rework the surface of the coastal plain and that soil materials go through a complete cycle from old upland surfaces, to thaw lakes, drained basins, and then back to conditions similar to the original is not consistent with our analysis of surficial materials, patterns of ground ice development, and geomorphology of lake basins on the Beaufort Coastal Plain. There simply is not enough ice in the original surficial deposits on most of the coastal plain to produce initial thaw lakes, and ice aggradation in the margins of drained basins is insufficient to allow thaw lake redevelopment. Furthermore, the rates at which lakes erode and drain, and at which ground ice develops, are too slow for the entire landscape to have been reworked by multiple cycles during the Holocene. Instead, we conceptualize a landscape that is altered by climatic changes, formation of large primary lakes in topographic depression, reworking of surficial materials in the primary lakes, development of integrated drainage networks, drainage, and differential development of ground ice within the drained basins. This sequential development of unique stages formed the conditions for the secondary development of small infilling ponds along the sandy margins of the basins and occasional thaw lakes in the ice-rich silty centers of old lacustrine basins. Lakes in much of northern Alaska conform to this sequence of development, while true thaw lakes are limited to areas with extremely ice-rich, silty deposits, such as near Barrow, the Colville River Delta, and the lower foothills of the Brooks Range.
 Field work in different parts of the Arctic Coastal Plain was funded by ConocoPhillips Alaska, Inc., and we appreciated the support of Caryn Rea, CPAI, in managing the projects. Additional support for mapping lake types and analysis was supported by National Science Foundation grants ARC-0454985 (to M.T.J.) and ARC-0454939 (to Y.S.). Allison Zusi-Cobb helped with cartography, and Betty Anderson helped with technical editing. We appreciate the dedicated reviews of Chris Burn and Ken Hinkel that helped improve the paper. Any opinions and findings expressed in this material are those of the authors and do not necessarily reflect the views of the CPAI and NSF.