Ice-wedge polygons are common, visually striking surficial manifestations of ground-ice in permafrost landscapes, particularly in Arctic regions with shallow permafrost and flat topography [1
]. Ice-wedge formation is initiated by the contraction and cracking of soils during winter freezing; then, during the thaw season, the incipient cracks fill with water that freezes the following winter and becomes incorporated in permafrost [3
]. Over centuries, wedge-shaped masses of ice enlarge through repeated seasonal cycles of cracking, thawing, and refreezing [5
]. Regional- and local-scale differences in climate, landscape position, and soil physical properties produce high variability in polygonal landforms and rates of ice aggradation across the landscape [6
]. Ice wedges can attain great size in silt-rich materials that are favorable for ground-ice aggradation, as found in yedoma landscapes of eastern Siberia and northwestern North America [7
], but develop at much slower rates in sandy material.
Alaska’s North Slope is underlain by continuous permafrost, and polygonal networks are nearly ubiquitous on the Arctic Coastal Plain and gently sloping, silt-rich uplands of the northern Arctic Foothills [8
]. Polygonal networks create complex microtopography and strong, meter-scale contrasts in vegetation, soil hydrology, ground-ice content, and thaw-settlement potential. Polygons evolve in concert with the growth (or thaw) of underlying ice-wedges, aggradation of ice beneath polygon centers, the development of drainage networks and thermal erosion, and regional climatic change. Polygons on the North Slope overwhelmingly reflect ice wedges that have developed during Holocene time, although very large, Pleistocene-aged wedges are preserved in the subsurface and occasionally exposed in yedoma soils [9
]. On the poorly drained Arctic Coastal Plain, the microtopography created by ice-wedge polygons strongly influences hydrologic flowpaths, spring runoff, and associated fluxes of dissolved nutrients [10
]. This microtopography also creates important habitat features for wildlife, including breeding waterbirds of high conservation concern [11
]. As a form of “patterned ground,” polygonal networks possess distinctive and diagnostic surficial attributes that facilitate detection and monitoring using high-resolution remote sensing.
Numerous reports of recent ice-wedge degradation have emerged from the North American Arctic, including the North Slope [12
], interior Alaska [15
], and western Canada [16
]. Degradation is initiated by thawing of the uppermost portions of ice wedges; the resultant subsidence forms small, flooded pits and troughs along the polygon margins that pock-mark the landscape and kill mesic-adapted vegetation [18
] (Figure 1
). Secondary impacts can affect large areas because pitting creates new hydrologic flowpaths that alter soil hydrology and the distribution of surface water [14
]. Over time, most pits become colonized by wetland vegetation and surface water extent declines due to the development of an organic mat [18
Ice-wedge degradation can be triggered by surface disturbance or inundation that alters the near-surface thermal regime, as observed near industrial infrastructure [13
], coastlines [20
], or after tundra fires [21
]. Widespread observations of recent ice-wedge degradation in undisturbed terrain, however, have implicated climatic warming as a triggering mechanism; for example, recent ice-wedge degradation near Prudhoe Bay, Alaska was triggered by unusually warm summers [13
]. Research conducted to date indicates that most degradation has been concentrated in the last two decades, which suggests that the warm summers of the late 20th century have exceeded the range of climatic variability on the North Slope over the last several centuries.
Here, we evaluate spatio-temporal patterns of ice-wedge degradation since the mid-20th century on undisturbed, residual upland surfaces in the coastal plain and foothills regions of Alaska’s North Slope using high-resolution imagery for three epochs: circa 1950, 1982, and 2012. Residual surfaces have not been modified by floodplain or lake-basin development and thus comprise the oldest surficial deposits in the region, where ice wedges have developed over long periods of time and virtually all surface water is restricted to thaw pits in polygonal ground. The broad longitudinal distribution of study areas captures upland terrain with differing surficial materials, landscape histories, and ground-ice content. To support retrospective analysis of high-resolution imagery with differing spectral and spatial resolutions, we quantified the extent of thaw pits by delineating potential waterbodies using a spectral thresholding approach, and manually deleted resultant “dark” objects that did not pertain to surface water in thaw pits. Although this introduced an element of subjectivity to the analysis, we believe this approach improved precision because we were able to exploit interpretive cues related to spatial relationships and landscape context that cannot be incorporated in a fully automated approach. Our analyses were also informed by field measurements of vegetation and soils, and detailed mapping of geomorphic units and microtopography at each study area.
The overarching objectives of our study were to (1) evaluate spatio-temporal patterns of ice-wedge degradation across the understudied central and western North Slope; and (2) evaluate patterns of correspondence between the observed changes and geomorphic and climatic gradients present across the region.
2. Materials and Methods
2.1. Study Area
We analyzed changes in the spatial extent of thaw pits over time at 11 study areas (15–50 km2
each) extending from near Point Lay on the Chukchi coastal plain, eastward across the National Petroleum Reserve–Alaska (NPRA) to near the Colville River (Figure 2
). The study areas are distributed across three major geomorphic units: (1) Tertiary-aged alluvial-marine deposits on the Chukchi coastal plain (“Chukchi sites”), (2) Pleistocene-aged eolian sand deposits of the central Beaufort coastal plain (“Beaufort sites”) [22
], and (3) Pleistocene-aged yedoma uplands of the Arctic foothills (“foothills sites”) [23
]. The study area locations were determined by the availability of coincident historical and modern imagery, and field plots sampled during 2010–2012. A strong climatic gradient exists across the region; summer air temperatures are higher and winter air temperatures lower in inland areas than coastal areas, although mean annual temperatures are nearly identical (−11.4–−11.1 °C) [24
] (Table 1
). Study areas on the coastal plain and foothills correspond to bioclimate subzones D and E of the Circumpolar Arctic Vegetation Map (CAVM), respectively [25
2.2. Data Sources
High-resolution imagery provides a record of thaw-pit extent for three epochs at each study area: circa 1950 (1948–1955; “early”), 1982 (1979–1985; “middle”), and 2012 (2009–2012; “late”) (Table 2
). The earliest imagery (1948–1949) comes from the U.S. Navy’s Barrow Area Reconnaissance program and provides panchromatic aerial photography at a 1-m resolution for four study areas; similar photography collected by the U.S. Air Force in 1955 provides coverage at a 2-m resolution for the remaining seven sites. The 1979–1985 imagery comes from the Alaska High Altitude Photography (AHAP) program and provides color-infrared (CIR) aerial photography at a 0.9–1.5-m resolution. Imagery for 2009–2012 comes from several commercial multi-spectral satellites with a 2-m spatial resolution; we sought multiple summer images for the late period, but the short snow-free season, high cloud frequency, and remote locations of the study sites greatly limited the availability of useful imagery. All imagery was acquired from late June to mid-August, and imagery acquisition dates spanned less than one month for all but one site. A digital elevation model (DEM) with 2-m posting derived from Interferometric Synthetic Aperture Radar (IFSAR) covered all study areas and we used it to inform the mapping of geomorphic units and microtopography.
2.3. Field Observations and Terrain Mapping
We used field data collected in 2010–2012 at a network of plots in NPRA to support mapping of geomorphic units, polygonal surface forms, and surface water extent (Section 2.4
, below). Field data were available for all but one study area (Kugachiak). At each plot, we collected data describing geomorphology, soils, vegetation, and disturbance following an ecological land survey (ELS) approach [27
]; these data were useful for classifying and mapping geomorphic units, polygonal landforms, and vegetation. Geomorphic units
refer to deposits of surficial materials that are the result of geomorphic, ecological, and periglacial processes and were classified according to a system developed for Alaska [28
]. Surface form
refers to the dominant microtopographic characteristics of ice-wedge polygons and was categorized according to Washburn (1980) [1
]. We collected vegetation species-cover and structure data to classify vegetation according to the Alaska Vegetation Classification [29
]. Finally, we recorded soil thaw depth, stratigraphy, and moisture regime at soil pits (40–50 cm deep). Stratigraphic data included the thickness of surface organic matter and soil physical characteristics; we also assigned a single simplified texture category (loamy, sandy, silty, or organic) to characterize the dominant soil texture and verify the geomorphic unit.
We used the field data to inform multi-parameter mapping of the study areas, including geomorphic units (alluvial-marine deposit, eolian sand, or yedoma) and polygon type (high-centered, low-centered, mixed high- and low-centered, or mixed pit–polygon complex). We digitized landscape patches by photo-interpreting the c. 2012 imagery for each study area at a scale of 1:4000 in ArcMap GIS software (ArcMap 10.5; ESRI, Redlands, CA, USA). The terrain mapping provided a means to stratify thaw pits by geomorphic unit and surface form and to assess differences in the timing and extent of thaw-pit development within and among study areas.
2.4. Surface Water Mapping
We first used ArcMap raster analysis tools to delineate waterbodies using the near-infrared (NIR) band in the c. 1982 and 2012 imagery; water strongly absorbs infrared radiation, whereas vegetated land reflects most incoming infrared. For each study area, we delineated all waterbodies through an iterative process in which we set a threshold NIR reflectance value in each image, and then assessed how well the threshold value distinguished water and land. After determining the best threshold value, we converted the resultant raster-based “water” pixels (or groups of adjacent pixels) to vector-based shapes. We used the same approach for the c. 1950 imagery using the single panchromatic band. Next, we used the terrain mapping to extract waterbodies that occurred in residual uplands (i.e., where surface water is restricted to thaw pits), and deleted waterbodies pertaining to lakes, ponds, drained-lake basins, and floodplains. We then visually reviewed the thaw-pit delineations and manually deleted dark shapes that were not thaw pits, such as shadowed bluffs and lake cutbanks. Finally, we summarized the spatial extent of thaw pits by geomorphic unit for each period and study area.
To evaluate the influence of interannual variability in surface water extent within a given observation period, we compared results using two images from summer 2010 (16 August; GeoEye-1) and 2011 (22 August; WorldView-2) that partially overlapped one study area (Kogosukruk). More extensive inter- and intra-annual comparisons were not feasible due to the sparse imagery archives for each of the study periods.
Using a retrospective remote-sensing approach, we found that ice-wedge degradation has driven increases in surface water extent since 1950 at 8 of 11 study areas spanning Alaska’s western and central North Slope (median change +130.3%). Although the recent, rapid increases in waterbody extent observed at several sites corroborate previous findings that ice-wedge degradation in the region was initiated by warm summers in the late 20th century, our finding that polygonal landscapes on the Chukchi coastal plain were already extensively degraded by 1950 lead us to conclude that transformation of polygonal networks in cold continuous permafrost is not solely constrained to the last few decades in northern Alaska, and likely elsewhere in the Arctic.
The use of surface water extent as a remotely sensed metric of permafrost dynamics has several strengths, mainly stemming from the distinctive spectral properties of surface water and its value as an indicator of underlying geomorphic conditions. We caution the remote sensing community, however, that confounding factors related to the specific geomorphic settings under study (e.g., low-versus high-centered polygons), and the potential influence of seasonal and inter-annual variability on surface water must be carefully considered. The multi-temporal, spectral thresholding approach presented here is best suited for flat landscapes with continuous permafrost, and requires that imagery come from the narrow, midsummer seasonal period.
Our findings provide context for interpreting and predicting the dynamics, climatic thresholds, and impacts of ice-wedge degradation in continuous permafrost of Alaska and elsewhere in the Arctic. While it has become clear that degradation of Holocene-aged ice wedges is underway in many parts of the Arctic, the broader ecosystem implications of these landscape changes remain poorly understood, with a range of potential consequences to ecosystem structure and function, as well as landscape sensitivities to industrial activity on the North Slope. Ice-wedge degradation appears to represent a rapid and conspicuous landscape response to extreme warm summers. Although successional processes tend to promote rapid vegetation recovery and paludification of thaw pits, the changes to landscape microtopography, soil hydrology, and hydrologic connectivity represent persistent impacts that are not reversible except on millennial timescales. Mismatches in the timing of ice-wedge degradation highlight the importance of landscape-scale differences in surficial geomorphology and regional climatic variability, and their role in modulating the response of Arctic environments to climate warming.