Satellite observations since 1979, taking the Arctic Circle as the region’s southern limit, indicate that Arctic near-surface temperatures have risen at almost four times the global average over recent decades [15,16]. Projections from the CMIP6 generation of climate models suggest that end-of-century Arctic warming will exceed earlier estimates-with some months warming by as much as 5°C-largely because precipitation-related processes were underrepresented in CMIP5 [17,18,19]. Several interacting mechanisms have been advanced to explain this phenomenon, known as Arctic Amplification (AA) [20,21].

The mechanism most frequently invoked is the ice-albedo feedback [22,23]: as rising temperatures shrink the sea-ice cover, the darker ocean surface that replaces it absorbs more solar radiation, reinforcing the initial warming. Across five models run under the SRES A1B scenario for the 21st century, albedo and polar temperature were found to co-vary in an essentially linear fashion [24], while the deposition of black carbon on the ice surface darkens it further and hastens melt [25]. A second driver is the intrusion of anomalously warm ocean water into the Arctic Basin [26]. Because the cold polar atmosphere responds disproportionately to incoming heat, these incursions-acting in concert with warmer winds, increasingly frequent heat waves [27], and wind-driven export of ice-have produced a year-on-year contraction of the sea-ice cover since the 1970s [28,29], leaving the stock of thick multiyear ice roughly 60% smaller than in the 1980s [30]. Coastal erosion is expected to intensify in step, with rates projected to grow 1.8- to 2.9-fold by the end of the century [27]-a sensitivity closely coupled to Arctic surface air temperature (SAT), which has climbed faster than the global mean relative to the historical baseline [31]. Warming also translates directly into global mean sea-level (GMSL) rise [32]: the rate accelerated from 1.1 mm/yr during 1900-1930 to 4.4 mm/yr during 2010-2015, with exposure concentrated in low-lying coastal zones [33]-and with it comes a growing flooding and erosion hazard for the infrastructure of those zones.

Permafrost degradation compounds these pressures. The same model ensembles that project rising temperatures also indicate rapid thaw [34]: active layers are deepening, and near-surface ground ice has diminished by up to 12% since 1850-and by roughly 20% in the most severely affected areas [35]. Field measurements corroborate the trend. At five monitoring sites in the western Russian Arctic, surface permafrost warmed across the board-from -8.0°C to -6.0°C at the northern sites and from -3.8°C to -1.9°C at the southern ones, converging toward -4.8°C and 0°C, respectively [36]. For cold-region infrastructure, U.S. military installations included, these trajectories are of immediate engineering consequence.

The U.S. Department of Defense (DoD) formally acknowledged the risk to its installations in Climate Change Adaptation and Resilience (2016), a directive that distributed responsibilities among DoD components, created a dedicated climate task force, and laid out a long-term adaptation roadmap [4]. By 2019 the Department was describing climate change as a matter of national security for its bases, singling out permafrost thaw and repeated sea-level-driven flooding as the leading hazards [5]. Yet a 2022 assessment concluded that two of every three of the United States’ 79 installations in the region were still exposed to climate impacts [6]. None of the six installations examined in that review had produced resilience-improvement plans, the standing directives notwithstanding: Army and Air Force leadership variously lacked familiarity with sustainability procedures, set aside the climate-risk recommendations of their own civil engineers, or made no use of the climate-projection tools available to them. Parallel interviews with Coast Guard officers and military engineers pointed to the same institutional pattern-engineers who lacked the means or mandate to act on sea-level rise and regarded it as a matter for higher authorities, a service that planned no further ahead than existing federal, state, and local guidance, and warming impacts that received attention only once infrastructure had already been damaged [7]. The 2024 DoD Arctic Strategy has since adopted a more action-oriented posture, allocating dedicated funding for permafrost mitigation at installations such as Pituffik Space Force Base in Greenland.

This study adopts a two-layer approach. First, a bibliometric analysis characterizes the evolution and thematic structure of the cold-region infrastructure literature between 2001 and 2026. Second, a focused comparative synthesis evaluates engineering practices across six nations to identify transferable lessons for U.S. Arctic infrastructure. Together, these approaches situate detailed engineering insights within a broader understanding of how the field has developed.

The focus on military installations is deliberate and methodological. Military infrastructure is not uniquely exposed to climate risk, but it is among the most comprehensively documented categories of cold-region infrastructure: defense agencies, inspector-general offices, governmental accountability organizations, and engineering commands publish unusually detailed, publicly available assessments of installation condition, climate vulnerability, maintenance cost, and adaptation activity-an evidence base with no equivalent for most civilian cold-region facilities. Military installations therefore serve here as a representative, information-rich lens through which the broader challenges of permafrost degradation, sea-level rise, and climate-resilient design can be examined and translated to civilian applications.

The scope is deliberately confined to the engineering of infrastructure and foundations in permafrost, Arctic, and cold-region environments under a changing climate; purely ecological, hydrological, or geophysical studies without an infrastructure dimension fall outside it. Within this scope, the review provides dedicated comparative treatment of six nations with substantial cold-region construction experience-the United States, Canada, Russia, Norway, Greenland (Denmark), and China-chosen for their documented engineering practice and their relevance as models for U.S. Arctic installations. Because the engineering challenges of frozen ground, thaw settlement, and climate-driven deterioration arise in both Arctic and high-elevation cold regions, the broader term “cold-region infrastructure” is adopted throughout this review, with “Arctic” reserved for contexts that are geographically Arctic; China accordingly enters the comparison as a high-elevation cold-region nation rather than an Arctic one.

The study employs a two-layer methodology. The first layer is a bibliometric analysis of the broader field. A topic search of the Web of Science Core Collection (SCI-EXPANDED, SSCI, and CPCI-S editions) was conducted, combining cold-region and permafrost terms with infrastructure and engineering terms and with climate-change and adaptation terms, restricted to English-language articles, reviews, and proceedings papers published between 2001 and 2026. After de-duplication and title/abstract screening against the inclusion criteria, the resulting corpus was analyzed in VOSviewer to map author-keyword co-occurrence and annual publication output characterizing the thematic organization of the field and growth. The complete search strategy-including the exact Boolean query as executed, field tags, and filters-is provided in Appendix S1 to ensure full replicability. The second layer is a focused narrative synthesis: from within the corpus, a curated subset of studies most directly addressing the four objectives and the six comparator nations was read in full and synthesized qualitatively, drawing additionally on authoritative defense, inspectorate, and governmental reports that document base conditions but are not indexed in Web of Science. Together, the two layers situate the detailed engineering review within a quantitative picture of how the field has evolved over the quarter-century.

Thawing permafrost undermines infrastructure through several distinct pathways. Because ice content, permafrost thickness, and ground temperature vary widely from site to site, engineering behavior in these terrains is difficult to predict [1,2]. Ground that is frozen at the time of construction provides firm, load-resistant support; once its ice melts, that bearing capacity is lost. Thaw additionally triggers thermokarst development and retrogressive slumping-processes that reshape the soil’s freeze-thaw regime [37] and can impose stresses approaching 300 kPa on the works above [1]. In the absence of established risk frameworks and the capacity for quantitative risk analysis, engineers have had to fall back on qualitative judgment, to the detriment of base performance [38]. Whole installations, moreover, stand on permafrost that is now degrading [39]: about 70% of all Arctic infrastructure rests on frozen ground, and roughly a third of that is sensitive to warming [40]. The form the permafrost takes also matters [41]. Sporadic permafrost binds soil constituents together, and its loss leaves irregular, hummocky terrain; continuous permafrost, by contrast, forms the foundation stratum beneath most Alaskan installations [42]. Some 15% of U.S. training-operations infrastructure and five DoD facilities in Alaska rest on such foundations, and as continuous permafrost degrades, the resulting subsidence endangers runways and the cold-dependent testing facilities that require subzero ground (Figure 3).

The damage is no longer hypothetical. In Russia, a fuel-storage tank ruptured as its foundation failed, releasing on the order of 150,000 barrels of oil toward the Arctic Ocean, and cumulative losses to transport links, industrial plants, and pipelines have reached hundreds of billions of dollars [40]. Severe thaw-driven runway flooding has been recorded at Thule (Pituffik) Air Force Base, while at Eielson AFB in Alaska an official report warned of complications for F-35A operations arising from thaw beneath the base [6], where hangars sit atop actively degrading permafrost with direct consequences for load-bearing capacity [44]. Climate considerations scarcely entered base design before the 1960s [39], and even once they did, long-term durability was rarely a design objective. In Alaskan and Canadian practice, the recurring problems are surface deformation, cracking and voids in foundations, and ponding of water along roadsides [1]. The bill for such damage may exceed $2 billion for the United States in the near term [39], rising to $7.3-14.5 billion by 2080 [45,46]. The Alaska-Canada Highway (ALCAN) illustrates a further, self-inflicted pathway: its builders stripped away the organic cover that had insulated the underlying permafrost [47], clearing forest and tundra and thereby accelerating thaw beneath a road that now serves military logistics.

Arctic warming feeds sea-level rise through accelerating ice loss, and the repercussions reach coastal infrastructure, maritime facilities, training ranges, and the supply chains that connect them [42]. A U.S. Naval War College paper had sketched these implications as early as 1990, but its claims were thinly substantiated and drew little response from civil engineers and planners. Today, 128 U.S. military bases-many of them in the Arctic-stand exposed to rising seas [10], and the DoD assesses two-thirds of its Arctic bases as facing recurrent flooding (Figure 4) [6]. A base need not suffer outright physical destruction to be lost to service [48]: without connectivity, transport links, power supply, and essential materials, it cannot operate.

Sea-level rise multiplies these vulnerabilities. When runways and berthing areas flood, aircraft, ships, and submarines can neither deploy nor be serviced-at high tide, sometimes not at all; budgets migrate from research and development into floodwalls and land reclamation; waterlogged grounds impede movement; and controlled training areas must be redesigned [42]. The domestic record already shows the pattern. Norfolk Naval Station in Virginia has recorded roughly 1.5 ft of sea-level rise over the past hundred years, and a comparable rise projected within the next 30-50 years would carry the base past its operational threshold [42]; higher tides there have already closed several low-lying docks, cutting capacity to a third. At Joint Base Langley-Eustis, water levels have climbed 14 inches since 1930, worsening flood frequency [5]; assessments of Naval Base Coronado likewise document impaired access, and flooding across the Hampton Roads area has at times cut off 29 military installations [49]. Arctic conditions sharpen the picture further: at Fort Greely, Alaska, an adjacent stream inundated close to five acres of the base to depths of 20 ft [6], eroding roadways throughout the installation. Recent pan-Arctic assessments sharpen this picture: roughly 60% of detected Arctic coastal infrastructure lies in low-lying zones, and a large share of coastal settlements is projected to be exposed to sea-level rise by 2100 [50], while the compound action of thaw subsidence, sea-level rise, and erosion may transform six to eight times more land than erosion alone, placing 40-65% of infrastructure in Alaskan coastal villages at risk by 2100 [51].

Although the field has historically depended on field monitoring and physical or numerical simulation-the COMSOL and SCAD analyses discussed above being representative-the bibliometric overlay in Section 2 identified a more recent methodological turn toward data-driven and scenario-coupled approaches. Two developments stand out. First, climate-scenario design, exemplified by the Norwegian practice of feeding general-circulation-model output into engineering models to size foundations against the warmest ground temperature expected over a structure’s service life rather than against historical conditions [12], reframes design as forward-looking. Second, machine-learning methods are increasingly used to predict permafrost behavior under uncertainty; for example, a Canadian framework couples long short-term memory networks with reanalysis data and regional climate projections to assess thaw threat along major rail and highway corridors [47]. These methods do not replace passive engineering or physical monitoring; rather, they confront the field’s central difficulty-designing fifty-year infrastructure against a moving climatic target-and constitute the most direct point of contact between cold-region engineering and the broader agenda of simulation-based optimization and predictive management in the built environment. For the United States, whose installations were largely designed against historical climate, adopting scenario-based and data-driven design represents a low-regret modernization of practice.

Published results illustrate that these approaches are already actionable. The Canadian LSTM framework, trained on reanalysis data and driven by regional climate projections, classified thaw threat along entire transportation corridors and found the Hudson Bay Railway, the Mackenzie Northern Railway, and the Inuvik-Tuktoyaktuk Highway all currently exposed [47]. In Svalbard, climate-scenario engineering models quantify how projected warming shortens a pile’s frozen embedment length and reduces its safe bearing load over the design life, and show that foundations dimensioned against projected end-of-life ground temperatures carry materially lower settlement risk than those dimensioned against historical climate alone, while yielding updated hazard maps for Longyearbyen [12,62].

Table 2 synthesizes the reviewed literature across the eight themes that structure the field, from the drivers of warming through its infrastructure impacts to the engineering, predictive, and comparative responses examined above. Alongside the bibliometric clusters of Section 2, it depicts a body of knowledge that is broad and maturing yet unevenly applied-well developed on passive cooling and foundations, thinner on Arctic-specific and predictive work-reaffirming that the principal deficit lies in translating available solutions into practice rather than in the absence of solutions.

Technology Readiness and Implementation Gaps: The U.S. adoption deficit can be stated precisely in the language the Department of Defense itself uses to manage technology maturation-the Technology Readiness Level (TRL) scale [92]. Most of the solutions reviewed here are not emerging technologies: thermosiphons and thermopiles, crushed-rock embankments, ventilated pile and spatial foundations, and coastal revetments have reached the highest readiness levels through sustained operational deployment in Canada, Russia, Norway, Greenland, and China, while PV-driven refrigeration and machine-learning-based predictive design occupy intermediate levels, validated in models and pilots but not yet standardized. The characteristic failure mode of the TRL pipeline-the “valley of death” between demonstrated technology and programmed, standardized adoption-therefore describes the U.S. position exactly: the gap lies not in research (TRL 1-4) nor in demonstration (TRL 5-7), which other nations have already supplied, but in the institutional transition to codified, funded, routine deployment (TRL 8-9). Framing the problem this way clarifies the remedy: not further invention, but the standard-setting, acquisition, and monitoring mechanisms addressed below.

Governance and Implementation Roadmap: The comparative review indicates that adaptation success depends on institutional integration as much as engineering innovation. A practical roadmap for U.S. Arctic and cold-region infrastructure would include: (1) standardized climate-scenario design requirements for new construction and major rehabilitation, on the Norwegian service-life model and the Russian SP 25.13330.2020 precedent; (2) a centralized permafrost-hazard and ground-ice database, following the Greenlandic geohazard-mapping example; (3) mandatory climate-risk assessment during infrastructure acquisition and siting; (4) expanded installation-level ground-temperature and settlement monitoring tied to the existing inspectorate reporting cycle; (5) dedicated adaptation funding streams insulated from reactive repair budgets; and (6) technology-transition pathways-pilot retrofits at the highest-risk installations such as Eielson and Pituffik, scaled through TRL-gated acquisition-linking research, demonstration, and implementation.

Economic Considerations: Although a full life-cycle cost comparison lies beyond the scope of this review, the available literature consistently indicates that proactive adaptation costs substantially less than projected climate damage. Estimated U.S. damages from permafrost thaw alone run upwards of $2 billion in the near term and $7.3-14.5 billion by 2080 (Section 3.1), whereas many of the reviewed measures can be incorporated at modest incremental cost during construction or scheduled rehabilitation: revetments proved the most cost-effective coastal defense against 2- to 500-year storm events in Beaufort Sea applications [85]; floodwalls require only periodic crack maintenance [86]; levees demand high upfront capital but predictable ~2% annual maintenance [83]; and dry floodproofing at Joint Base Langley-Eustis cut sandbag deployment by 70% with simple door dams and flood-projection models [91]. Current dedicated spending remains small against the exposure-the 2024 DoD Arctic Strategy allocates roughly $7 million for permafrost mitigation at Pituffik [63]. For the foundation and refrigeration classes (Multipoint Foundations, VCRS/ARS), published head-to-head costs against conventional retrofits are lacking, and comprehensive cost-benefit analysis across passive cooling, foundation retrofit, and coastal-protection strategies under multiple climate scenarios stands as a priority for future research.

Taken together, these strategies form a complementary system rather than isolated solutions. Effective adaptation in cold-region infrastructure depends on integrating passive cooling, foundation design, coastal protection, and predictive modeling into a unified framework responsive to both current and projected climatic conditions.