Alaskan Permafrost - Queries and Joins
Travis Zalesky
As part of UA GIST 602B
Travis Zalesky
As part of UA GIST 602B
Figure 1. Alaskan permafrost classification study area. Permafrost classification from the Circum-Arctic Map of Permafrost and Ground-Ice Conditions, Version 2.
Melting permafrost, particularly thermokarst, is an emerging concern for residents and communities in and around the Arctic. Largely driven by anthropogenic climate change, it poses a wide variety of risks, not the least of which is damage to infrastructure due to ground subsidence and rapid erosion. How can data be leveraged to inform us which areas, and which populations, are at the highest risk of infrastructure loss or damage in a changing climate? Are certain communities or populations at a disproportionate risk? Data exploration tools such as queries and joins can be used to answer a wide variety of research questions. Herein, I will describe the function and use case of a variety of common data exploration tools for both geospatial and non-spatial data, as I explore the risks and potential consequences of melting Alaskan permafrost.
Permafrost is a layer of frozen ground that has remained continuously frozen for at least 2-years. Permafrost can vary in thickness from less than a meter up to 1.5 kilometers, and is often (though not always) covered by a thin layer of soil which undergoes an annual freeze-thaw cycle(s), known as the "active layer" (National Snow and Ice Data Center, 2024; United States Permafrost Association, 2021). The active layer, combined with any additional mass overtopping the permafrost, is collectively known as the "overburden". Permafrost is a major feature of the polar and subpolar regions of the Northern Hemisphere, covering as much as 25% of the total land mass N of the Equator (National Snow and Ice Data Center, 2024). Although permafrost need only be frozen continuously for 2-years, much of the world's permafrost, particularly in and around the Arctic, has been continuously frozen for thousands of years (Turetsky et al., 2019).
Although there are still many unresolved questions about how climate change will impact permafrost, it is clear that, globally, permafrost is receding. Melting permafrost has many, widespread, implications, from exacerbating climate change due to release of sequestered carbon into the atmosphere, to human health impacts from the liberation of pollutants, or possibly even novel human pathogens (Langer et al., 2023; Miner et al., 2021; Turetsky et al., 2019). However, one of the most tangible, immediate impacts of melting permafrost to individuals and communities is the very real potential for infrastructural damage (Hjort et al., 2018, 2022; Melvin et al., 2017; Suter et al., 2019).
Receding permafrost can be broadly characterized into two modes of melting. Historically, permafrost melt have been modeled as a slow, gradual process of incremental thinning of the frozen layer, from the top down (i.e., active layer thickening). However, recent research has shown that there is another, more rapid and dramatic form of permafrost melting, known as thermokarst (Olefeldt et al., 2016). Thermokarst melting is defined by distinctive, rapid, land subsidence and erosion patterns, that often take the form of shallow lakes or depressions (Fig. 2; Matuszewski, 2017). More generally, thermokarst can also cause landslides or slumps in hilly or mountainous regions, and can contribute to massive erosion along riverbanks. Due to the rapid, and often dramatic, nature of thermokarst melting, it is this mode of melting in particular that has the potential for devastating infrastructural damage to the built-environment. Perhaps counterintuitively, it is the permafrost with the highest ice content that is most at risk for thermokarst thawing (Hong et al., 2014).
Figure 2. Development of a thermokarst lake. From "Sediment-biogeochemical characterization of a thermokarst lake basin in Central Alaska" Matuszewski, 2017 (Figure 3).
Furthermore, there is an emerging body of work highlighting melting permafrost as an environmental justice issue. Much of the polar and subpolar regions of North America are the ancestral homes of Native communities, which rely on stable ice conditions for their subsistence and ways of life (Kovacs, 2019). Above and beyond the risk of loss of culture, heritage, and history to Native communities in the face of a rapidly changing environment, due to the remoteness of many of these communities, they may face disproportionate difficulty in acquiring the resources necessary to mitigate and repair structural damage caused by melting permafrost.
Using a case study in Alaska, I will demonstrate the use of various geospatial data science techniques while also endeavoring to answer various questions relating to the impacts of melting permafrost on communities and infrastructure (Fig. 1).
Attribute queries are a simple and straightforward data exploration technique. In brief, it is the use of keywords to subset and summarize data within a database or data table with some common feature. It is a broadly applicable technique, and not strictly geospatial, although it is commonly used in geospatial datasets. There are many means of querying a dataset, although the most notable and widely applied method, including applications in Esri ArcGIS Pro, is through use of Standard Query Language (SQL). Attribute queries can be used to answer questions such as, (1) How much land in Alaska is highly vulnerable to melting permafrost hazards?, and (2) Which are the largest communities (population) in Alaska?
Spatial queries are a geospatial subset of attribute query. In a spatial query, the attribute(s) that you are using to subset your data is based on location. Geospatial queries can be applied to any point, line, or polygon features, and some typical geospatial queries may take the form of, points within a polygon, lines intersecting a line, polygons containing a point, or polygons adjacent to a polygon, etc., etc. Spatial queries can be used to answer questions such as, (1) How many communities are within Native Alaskan designated census areas?, and (2) How many cities are at high risk of infrastructural damage from melting permafrost?
Spatial joins are a means of combining datasets based on some location data. They link attributes from two or more tables together based on their spatial relationship. They can take the form of a one-to-one or one-to-many relationship, depending on if one or more of the features to be joined are coincident with the target feature (see spatial queries). Spatial joins can be used to answer questions such as, (1) How many people reside in communities that are at risk of infrastructural damage from melting permafrost?, and (2) How many residents live within Native Alaskan designated census areas?
Finally, using a combination of the techniques described above, we can answer more complex questions, such as, (1) How many road miles are vulnerable to damage due to melting permafrost?, and (2) Are residents within Native Alaskan communities at a disproportionate risk of infrastructural damage due to melting permafrost?
All analysis and figure generation was performed in Esri ArcGIS Pro v3.2.2, using the NAD 1983 Alaska Albers Equal Area Conic projection (standard parallels, 55N & 65N; EPSG:3338). This projection was chosen due to the large E-W extent, as well as the far N. latitudes of the study area. While this projection is known to distort distance and shape, it preserves area, and is the most appropriate compromise projection for the whole of Alaska.
Data was obtained from a variety of publically available sources. Permafrost data was acquired from the National Snow and Ice Data Center's, Circum-Arctic Map of Permafrost and Ground-Ice Conditions, Version 2 (Brown et al., 2002). All other data was acquired through ArcGIS Online. Data layer sources and metadata are summarized in Table 1.
The permafrost classification layer was clipped to the extent of Alaska, which was itself selected from the US States layer. Further, ocean classification was selected from the clipped permafrost layer (attribute query), and was removed.
A new quantitative risk level field was added to the permafrost attribute table, which was determined from the qualitative permafrost classification. While there are many unknown factors that contribute to the risk of thermokarst thawing in permafrost, and there is a large degree of uncertainty regarding risk classification, even amongst leading experts, it is generally accepted that ground ice content is the most significant factor effecting thermokarst thawing (Hong et al., 2014; Olefeldt et al., 2016). Therefore, I was able to calculate the overall risk based on the sum of permafrost extent, ground ice content, and landform risk factors, a modified method based on the Permafrost Settlement Hazard Index, developed by Hong, et al. (2014). Individual risk factors were ranked on a 1 to 5 scale, as detailed in Table 2.
The Department of Commerce, Community, and Economic Development (DCCED), Certified Population layer (henceforth referred to as "communities" layer/data) contained data from multiple census years for each community, creating a redundant dataset for the purposes of this analysis. A new layer with the most recent population data was initially selected using the select tool, with the "Data Year" field equal to 2022 (an attribute query). A few communities did not have data available for 2022, therefore the communities that did not intersect with the new 2022 population data layer (a spatial query) were selected using the select by location tool. Forty-seven records were returned, representing 7 communities. The attribute table was explored, and only the most recently available population data for each of the 7 communities was selected. This filtered selection was then saved to a new layer, and the missing communities were merged with the larger 2020 population data layer using the merge tool.
The regions at high risk of infrastructural damage due to melting permafrost (i.e., risk level 9 or 10) were selected and saved to a new layer using the "Select" Analysis tool (Fig. 3). The total area at high risk was calculated as the sum of the autopopulated polygon "Shape Area" fields (Km2) field, using the summary statistics provided with the "Create Chart" histogram feature.
The largest cities in Alaska were subset and saved to a new layer using the "Select" Analysis tool, by querying communities with a population equal to or greater than 10,000 citizens (Fig. 4).
Figure 3. Workflow demonstration, including SQL query, for selection of areas at high risk for thermokarst permafrost melting.
Figure 4. Workflow demonstration, including SQL query, for selection of large cities.
Native American communities were identified from the communities layer by using a spatial query to select communities that were within Alaskan Native designated census areas, using the select by location tool (Fig. 5).
High risk cities were identified from the cities layer using a spatial query, selecting communities that were within the high risk permafrost layer (Fig. 6).
Finally, communities that were both Native Alaskan, and at high risk were identified by the intersection of the two respective layers (Fig. 7). Each selection was saved to a new feature layer.
Figure 5. Workflow demonstration for selection of Native Alaskan communities that lie within Alaskan Native designated census areas.
Figure 6. Workflow demonstration for selection of communities that lie within areas at high risk for thermokarst permafrost melting.
Figure 7. Workflow demonstration for selection of communities that are both high risk, and Native Alaskan.
The total population living within high risk areas was calculated by first dissolving the high risk permafrost polygons, creating a single, large, multi-part polygon. The population data was then joined to the high risk permafrost layer using a one-to-one spatial join (Fig. 8). The number of communities within the high risk permafrost polygon was counted, and their populations were summed.
The population data was joined to the Native Alaskan designated census area layer using a one-to-one spatial join (Fig. 9). The number of communities within each census area was counted, and their populations were summed. The total number of residents was calculated using the summary statistics provided with the "Create Chart" histogram feature.
Figure 8. Workflow demonstration for the spatial joining of population data to the high risk permafrost layer.
Figure 9. Workflow demonstration for the spatial joining of population data to the Native Alaskan designated census areas layer.
The dissolved high risk permafrost layer was combined with the roadways layer using the intersect tool. The total road length was calculated using the autopopulated shapelength field (Fig. 10).
The Native Alaskan designated census areas were combined with the dissolved high risk permafrost layer using the intersect tool (Fig. 11). This process sub-sections the two input polygons, returning only those areas that are both Native Alaskan and at high risk. The population per high risk census area, was calculated using a one-to-one spatial join, as before (Fig. 12). The number of at risk Native Alaskans was compared to the total Native Alaskan population, and was expressed as a percentage of at risk population. Likewise, the non-Native percentage at risk, and the total Alaskan population percentage at risk were calculated, and the population statistics were compared.
Figure 10. Workflow demonstration for the intersection of roadways with the high risk permafrost layer.
Figure 11. Workflow demonstration for the intersection of Native Alaskan designated census areas to the high risk permafrost layer.
Figure 12. Workflow demonstration for the spatial joining of population data to the high risk Native Alaskan designated census areas.
The overall risk classification ranged from 10 (highest risk) to 0 (not applicable). The majority of the state was found to be at moderate to high risk of infrastructural damage due to melting permafrost (5 to 10). However, the Aleutian Islands, as well as the coastal region around the Gulf of Alaska, was identified as no risk (Fig. 13).
Figure 13. Potential risk of infrastructural damage from melting permafrost calculated as the sum of risk factors derived from National Snow and Ice Data Center permafrost categorization. Risk factor classification derived from a simplified Permafrost Settlement Hazard Index (Hong et al., 2014). Amorphous gray regions in S. Central Alaska are characterized by small and complex polygons, which are primarily low, or no risk.
Figure 14. Areas at high risk of thermokarst melting and large cities, each selected via attribute queries.
The total area of land at high risk of infrastructural damage was found to be 640,211 Km2, covering more than 1/3rd of the state. Fourteen large cities were identified with a population >10,000. Several of these communities are suburbs or municipalities clustering around either Fairbanks, or Anchorage. Aggregating these closely aligned clusters results in 8 distinct population centers (Fig. 14).
Of 475 total communities with population data, 129 were identified as high risk, lying within a region at high risk of thermokarst permafrost melting. Additionally, 290 communities were identified as lying within Native Alaskan designated census areas. There were 97 communities identified as both Native Alaskan and high risk (Fig 15).
Figure 15. Native communities and high risk non-native communities across Alaska.
There were a total of 92,117 residents living in the 129 communities within the areas at high risk for thermokarst melting. Additionally, there were a total of 221 Native Alaskan designated census areas, in which 489,819 people reside. Native Alaskan populations ranged from 0 residents to 135,359 residents in Chickaloon, with a median population of 202.5 residents per census area.
More than 2,100 roads (or road segments) were identified as at high risk, a total road length of 735 miles (Fig. 16).
The total Alaskan resident population in 2022, according to the DCCED was 1.2 million. This is a higher number than the US Census estimate of 734 thousand, although it is not known how the two census methodologies differ. There was a total of 490 thousand Alaskans residing within the boundaries of an Alaskan Native designated census area. While it is assumed that the majority of those residents are, themselves, Alaskan Natives, this may not always be the case. The remaining 700 thousand residents are assumed to be non-native, although again, this may not be entirely true.
Figure 16. Major Alaskan roadways at high risk to infrastructural damage due to melting thermokarst permafrost.
The total population at high risk was calculated as 92 thousand residents, with 62 thousand Alaskan Natives and 30 thousand non-native. The total number of Alaskan residents at risk was approximately 8%. However, nearly 13% of the Alaskan Native population was found to reside within high risk areas, as opposed to only 4% of the non-native population. A full breakdown of population statistics is provided in Table 3.
Queries and joins are powerful data exploration techniques for both spatial and non-spatial data sets. However, the unique nature of geospatial datasets allows for additional queries to be made, based on inherent spatial relationships, such as contains, intersects, or adjacency, etc. These techniques, individually, and in concert, allowed us to explore various datasets related to melting Alaskan permafrost, and identify where people are most vulnerable to resulting infrastructural damage.
Perhaps counterintuitively, areas with thin and sporadic permafrost are at relatively low risk from melting permafrost. It is the regions with a large and continuous ground ice presence that are in the greatest danger of experiencing thermokarst melting. There is a wide variety of risks associated with melting permafrost, however our focus was on infrastructural damage, which is an immediate and tangible outcome effecting communities and individuals in high risk areas.
Using a custom risk classification system that was developed from a simplified Permafrost Settlement Hazard Index (Hong, et al. 2014), I found that over 1/3 of Alaska is at high risk due to thermokarst melting, particularly in the North, and on the Steward Peninsula, as well as a substantial portion of the Central inland region (Fig. 14). This is largely in agreement with previously published assessments, although there remains a substantial degree of uncertainty in the field (Hong et al., 2014; Olefeldt et al., 2016). Using an attribute query, large Alaskan cities were identified, with a population greater than 10,000 residents. Although some larger cities were broken down into smaller census counts, such as the municipalities surrounding Anchorage, there were 14 large cities identified, of which only one, Utqiagvik/North Slope Borough (10,746 residents), was on a high risk permafrost layer (Fig. 14).
Additional spatial queries were performed to identify the intersections between population centers, Native Alaskan designated census areas, and high risk thermokarst areas (Fig. 15). The majority of Alaskan communities are not at high risk from thermokarst, with many communities in the South and surrounding the Gulf of Alaska at little to no risk. However, more than 1/3 of Alaskan Native communities are at risk, compared to only 27% of all communities. The disproportional risk to Alaskan Native communities is further emphasized by the fact that Native communities accounted for 61% of all DCCED community census data, with only 17% of non-native communities in high risk areas.
Joining datasets on spatial relationships allowed further exploration and summarization of population data and was a convenient means of calculating simple summary statistics.
The final analysis found that nearly 13% of Native Alaskans may be a high risk to loss of personal or community property due to thermokarst melting. This is nearly twice the risk borne by non-native residents. This finding supports claims by Native Alaskan and environmental justice advocacy groups, that Native Alaskans are disproportionately impacted by thermokarst melting, a direct result of anthropogenic climate change.
Additionally, hundreds of miles of major Alaskan roadways are at high risk, an issue effecting both Native and non-native populations, particularly those in rural communities. Due to the known distance distortions inherent with the Albers Equal Area Conic projection, this analysis should be considered an estimate of at risk road miles, used to inform subsequent analysis. Interested parties are advised to identify local road segments, or study areas of interest and calculate road segment length using an appropriate regional map projection such as Universal Transverse Mercator, or a State Planes projection.
Due largely to the uncertainty surrounding the factors influencing thermokarst melting, these results should be considered exploratory, rather than authoritative. Thermokarst melting is a complex and poorly understood process, and although the overall risk assessment is largely in agreement with prior research, it is based on a limited dataset, with a much simplified risk model. Additionally, the "high risk" threshold level used in this analysis is largely arbitrary. Reducing the threshold risk level would massively increase the population and communities considered at risk (Fig. 12). Moreover, no attempt was made to classify future risk under climate change models. While ground ice content is considered to be the greatest single factor influencing thermokarst melting, other factors such as air and ground temperature are also thought to be significant (Hong et al., 2014). However, thermokarst melting is a clear and present danger to communities and infrastructure, and one that is likely to be exacerbated in the future, proportional to the level of average atmospheric warming. Thermokarst melting is a highly dynamic process, and unfortunately, it can not yet be easily predicted on small scales. It will continue to impact communities in and around the Arctic for the foreseeable future, and it is recommended that individuals, as well as local and federal governments, understand their level of risk and prepare for the impacts of melting permafrost.
Brown, J., Ferrians, O. J., Heginbottom, J. A., & Melnikov, E. S. (2002). Circum-Arctic Map of Permafrost and Ground-Ice Conditions, Version 2. National Snow and Ice Data Center. https://doi.org/10.7265/skbg-kf16
Hjort, J., Karjalainen, O., Aalto, J., Westermann, S., Romanovsky, V. E., Nelson, F. E., Etzelmüller, B., & Luoto, M. (2018). Degrading permafrost puts Arctic infrastructure at risk by mid-century. Nature Communications, 9(1), 5147. https://doi.org/10.1038/s41467-018-07557-4
Hjort, J., Streletskiy, D., Doré, G., Wu, Q., Bjella, K., & Luoto, M. (2022). Impacts of permafrost degradation on infrastructure. Nature Reviews Earth & Environment, 3(1), 24–38. https://doi.org/10.1038/s43017-021-00247-8
Hong, E., Perkins, R., & Trainor, S. (2014). Thaw Settlement Hazard of Permafrost Related to Climate Warming in Alaska. Arctic, 67(1), 93–103. JSTOR.
Kovacs, E. (2019). Environmental Justice Disparities in Alaska Native Communities. D.U.Quark, 4(1), 40–49.
Langer, M., von Deimling, T. S., Westermann, S., Rolph, R., Rutte, R., Antonova, S., Rachold, V., Schultz, M., Oehme, A., & Grosse, G. (2023). Thawing permafrost poses environmental threat to thousands of sites with legacy industrial contamination. Nature Communications, 14(1), 1721. https://doi.org/10.1038/s41467-023-37276-4
Melvin, A. M., Larsen, P., Boehlert, B., Neumann, J. E., Chinowsky, P., Espinet, X., Martinich, J., Baumann, M. S., Rennels, L., Bothner, A., Nicolsky, D. J., & Marchenko, S. S. (2017). Climate change damages to Alaska public infrastructure and the economics of proactive adaptation. Proceedings of the National Academy of Sciences, 114(2), E122–E131. https://doi.org/10.1073/pnas.1611056113
Miner, K. R., D’Andrilli, J., Mackelprang, R., Edwards, A., Malaska, M. J., Waldrop, M. P., & Miller, C. E. (2021). Emergent biogeochemical risks from Arctic permafrost degradation. Nature Climate Change, 11(10), 809–819. https://doi.org/10.1038/s41558-021-01162-y
Matuszewski, F. (2017). Sediment-biogeochemical characterization of a thermokarst lake basin in Central Alaska. https://api.semanticscholar.org/CorpusID:134218025
National Snow and Ice Data Center. (2024). Frozen Ground & Permafrost [University Website]. https://nsidc.org/learn/parts-cryosphere/frozen-ground-permafrost
Olefeldt, D., Goswami, S., Grosse, G., Hayes, D., Hugelius, G., Kuhry, P., McGuire, A. D., Romanovsky, V. E., Sannel, A. B. K., Schuur, E. A. G., & Turetsky, M. R. (2016). Circumpolar distribution and carbon storage of thermokarst landscapes. Nature Communications, 7(1), 13043. https://doi.org/10.1038/ncomms13043
Suter, L., Streletskiy, D., & Shiklomanov, N. (2019). Assessment of the cost of climate change impacts on critical infrastructure in the circumpolar Arctic. Polar Geography, 42(4), 267–286. https://doi.org/10.1080/1088937X.2019.1686082
Turetsky, M. R., Abbott, B. W., Jones, M. C., Walter Anthony, K., Olefeldt, D., Schuur, E. A. G., Koven, C., McGuire, A. D., Grosse, G., Kuhry, P., Hugelius, G., Lawrence, D. M., Gibson, C., & Sannel, A. B. K. (2019). Permafrost collapse is accelerating carbon release. Nature, 569(7754), 32–34. https://doi.org/10.1038/d41586-019-01313-4
United States Permafrost Association. (2021). Permafrost Glossary [Nonprofit Website]. https://www.uspermafrost.org/permafrost-glossary